# Burna AI > The safety and data quality platform for clinical research, from early phase trials through postmarket surveillance. Burna AI performs automated adverse event grading and attribution in real time, providing clear rationale and audit-ready evidence, including generation of IRB, sponsor and FDA reports. It delivers CTCAE v5 and v6 grades and MedDRA LLT codes on every encounter, enabling standardization, cross-site consistency and accuracy of clinical research data. Every grade is bound to the CTCAE criterion and the line of the medical record that produced it, and a clinician reviews and signs it: AI suggests, clinicians decide. Depth is in oncology, where the grading standard is hardest. Below are published articles, the product FAQ, and structured FAQ entries from site content (Velite) in plain text. ## Blog - [Model identifier on every grade, automatic organ-system routing, and mobile sign-in for institutional accounts](https://burna.ai/blog/2026-W01-weekly-update): First Friday update of 2026. Every grade now writes the language-model identifier into its audit trail, every event routes to a system organ class automatically, and the mobile application supports password sign-in for institutions that require it. - [Mobile CTCAE grading, a recent-cases dashboard, and an authentication flow that respects an interrupted clinician](https://burna.ai/blog/2026-W03-weekly-update): CTCAE grading now works fully on the mobile app. The recent-cases dashboard surfaces severe and pending events. Authentication state survives interruptions, app backgrounding, and device restarts. - [Multi-event grading in one session, an instant CTCAE term search, and inter-rater analytics inside the platform](https://burna.ai/blog/2026-W04-weekly-update): Coordinators can now grade up to ten adverse events for a patient in a single session, search the full CTCAE term set with sub-second response, and view inter-rater agreement statistics without exporting data. - [Workflow resumption after failure, one-click clinical data export, and mobile navigation that respects the floor](https://burna.ai/blog/2026-W05-weekly-update): Failed multi-step workflows now resume from the last successful step rather than from scratch. Clinical data exports compile in one click. The mobile navigation matches the rhythm of clinical work. - [Patient-level adverse event aggregation, encounter-level review and edit, and a unified grading substrate across web and mobile](https://burna.ai/blog/2026-W06-weekly-update): The patient view now aggregates adverse events across encounters. The encounter view supports accept, edit, and add of AI suggestions with a tracked review status. Web and mobile now share one type system and one filter language for graded events. - [Blind grading, designated grader roles, and grade definitions one click from the grading form](https://burna.ai/blog/2026-W07-weekly-update): The platform now supports blind grading with bulk case seeding and assignment, designated grader roles for quality programs, and inline grade definition reference during clinician review. - [Automated adjudication when graders disagree, real-time grading analytics, and a deliberate removal that mattered](https://burna.ai/blog/2026-W08-weekly-update): Grade conflicts between independent raters now route automatically to a designated adjudicator pool, resolve under e-signature, and write to an auditable trail. Analytics dashboards now reflect live operations. And one feature came out of the product because it could not be done responsibly. - [End-to-end EHR integration, submission-gated blinding, and MRN-based patient search](https://burna.ai/blog/2026-W09-weekly-update): Coordinators can now authenticate directly with Epic, Oracle Health, or AthenaHealth and pull patient data into the grading workflow without leaving the platform. Submission-gated blinding hardens the multi-rater workflow. Patient search now spans Medical Record Numbers. - [Direct patient import from the EHR, workflow resumption from the failed step, and automatic alerts for Grade 3 events](https://burna.ai/blog/2026-W10-weekly-update): Patient demographics now import directly from Epic, Oracle Health, or AthenaHealth. Failed multi-step workflows resume from the failed step rather than from scratch. Patient-reported Grade 3 events now route to provider notifications automatically. - [EHR encounter import in one step, batch adjudication, and 150 CTCAE synonym terms for trial documentation](https://burna.ai/blog/2026-W11-weekly-update): Clinical notes now import directly from the EHR into the grading workflow. Grade disagreements resolve in batch rather than one at a time. The synonym mapper covers immune-related adverse events and CAR-T-specific patterns. - [Comorbidity-aware grading, copy-forward detection, self-improving terminology, and 42 oncology regimen profiles](https://burna.ai/blog/2026-W11b-weekly-update): CTCAE grading now considers patient comorbidities, detects copy-forward documentation patterns, learns institutional vocabulary, and ships with 42 prebuilt oncology regimen profiles for multi-drug attribution. - [Connected analysis in one pass, clinical-grade AE report PDFs, and real-time workflow streaming](https://burna.ai/blog/2026-W12-weekly-update): CTCAE grading, multi-drug attribution, and clinical evidence analysis now run as one connected pipeline. The platform generates branded, audit-ready adverse event report PDFs. Each workflow step streams progress as it runs. - [Regulatory report generation, automatic visit titles from clinical notes, and study site management](https://burna.ai/blog/2026-W13-weekly-update): Coordinators can now generate FDA MedWatch, IRB, and Sponsor SAE reports directly from graded adverse events. Visit titles and excerpts generate automatically from clinical notes. Study sites are now a first-class concept in the platform. - [Institution-specific grading preferences, expanded EHR resource coverage, and severity-ordered adverse event review](https://burna.ai/blog/2026-W14-weekly-update): Organizations can now configure grading rules specific to their institution. EHR integration coverage expanded materially. Adverse events now order by grade in descending severity, with high-severity events expanded by default. - [Knowledge base ingestion for context-aware grading, protocol-anchored multi-drug attribution, and a patient mobile app](https://burna.ai/blog/2026-W15-weekly-update): Upload protocols, investigator brochures, and institutional guidelines, and the AI weighs them during CTCAE grading. Multi-drug attribution is now anchored to protocol-specific expectations. Patients can self-report symptoms from a dedicated mobile app. - [Patient-reported symptoms now run through the full CTCAE grading pipeline](https://burna.ai/blog/2026-W16-weekly-update): Patient PRO-CTCAE submissions now route through the same clinician-grade CTCAE workflow as provider-entered events, with clinician sign-off under a 21 CFR Part 11 e-signature. Plus token reissuance and mobile app polish. - [Workflow streaming on a native protocol, a new burna.ai front door, and a real staging environment](https://burna.ai/blog/2026-W18-weekly-update): The real-time streaming pipeline behind every AI workflow moved onto a native workflow protocol. The new burna.ai marketing site launched. Staging environment now lives for QA validation before production rollouts. - [Trial protocols become searchable: protocol ingestion threads protocol-specific evidence into every CTCAE grade](https://burna.ai/blog/2026-W19-weekly-update): Upload an oncology trial protocol and the platform parses it into structured metadata, embeds the document for semantic retrieval, and threads protocol-specific evidence into every CTCAE grading run. Plus a mobile-responsive provider portal and a passwordless demo flow. - [Comorbidity-aware attribution, real-time PRO submission freshness, and a trial UI in seven languages](https://burna.ai/blog/2026-W20-weekly-update): Attribution reasoning now factors patient comorbidities. The provider roster surfaces PRO submission freshness directly. The trial management UI ships internationalized in seven languages with inline term definitions. - [Patient self-report becomes a copy-forward signal, the recording pipeline gets a safety valve, and the architecture ships as nine animated figures](https://burna.ai/blog/2026-W22-weekly-update): Copy-forward detection now reads patient PROs and clinician grades as one combined evidence stream. The recording flow gains a 45-second timeout and a codec fallback chain. Nine animated architecture figures replace static videos across the audience landings. - [CTCAE v6.0 Migration Guide, Part 1: What Actually Changed](https://burna.ai/blog/ctcae-v6-migration-01-overview): A coordinator's overview of the v5.0 to v6.0 transition. What changed, what didn't, and how to think about retraining your team. ## Product FAQ (/faqs) Full text (machine-readable): [https://burna.ai/llms.mdx/faqs](https://burna.ai/llms.mdx/faqs) Human page: [FAQ — Burna AI](https://burna.ai/faqs) ## What Burna is, and what it is not. ### What is Burna in one sentence? Burna is the safety and data quality platform for clinical research, from early phase trials through postmarket surveillance. AI suggests, clinicians decide. ### Do you replace clinicians? No. AI suggests, clinicians decide. Human-in-the-loop, always. Every CTCAE grade is reviewed and signed by a clinician before it leaves the platform. ### How is this different from a single LLM with retrieval? A single LLM with retrieval is one model and one prompt with a citation step bolted on. The Burna engine is twelve specialized agents in a cascading constraint pipeline. Each agent's output bounds the valid output space of every downstream agent, terminating at a citation-bound CTCAE grade that a clinician signs. The engine structurally cannot produce a grade outside the defined CTCAE set, cannot skip citation, and cannot contradict its own upstream findings. Constraint at the architecture level, not the prompt level. ### Could someone build this in a weekend? A weekend project produces a single classifier or a single LLM call. The Burna pipeline spans twelve agent categories with cascading constraints between them. Two patents filed on the cascade. ### Will Veeva, Medidata, or another platform vendor add this as a feature? They operate downstream of the attribution decision, in safety case management and EDC. Burna operates upstream, at the point where the investigator is making the attribution decision. The white space is structural, not feature-level. Burna integrates with Veeva Vault EDC, Medidata Rave, and other downstream systems. ## Validation and evidence. ### How accurate is the grading? Strong agreement with expert clinicians in internal validation (97%+ agreement). Full statistical analysis plans and current figures are shared under NDA during pilot scoping. Headline accuracy percentages live in peer-reviewed papers, not on marketing pages. ### What validation studies are in flight? Mayo Clinic Platform Accelerate 1,200-chart accuracy study (2026 cohort); paired-grader prospective workflow study; second-site workflow validation; postmarket pharmacovigilance pilot scoped for 2027. ### Who is advising the work? Clinical advisors include oncology faculty at Duke and other academic centers. These are advisor relationships, not institutional partnerships. We name advisors only where they have given explicit consent. ### What happens if a design partner pilot does not hit its endpoints? For academic medical center pilots in the 2026 design partner cohort, the institution keeps the deployment, training materials, change-management playbook, implementation lead's work product, and the co-authored manuscript. Risk reversal applies on those pilots. ## Deployment, EHRs, and EDCs. ### How long does integration actually take? Two to three weeks from kickoff to first signed grade is the working target. Your IT provisions one OAuth client and one FHIR endpoint. Burna's engineering handles the rest. ### Do you support our EHR? SMART on FHIR is the integration layer. Burna runs inside Epic, Oracle Health (Cerner), Athenahealth, Allscripts, eClinicalWorks, MEDITECH, Dedalus Orbis (France, Italy, Spain), AGFA Orbis (Germany, Benelux), and regional APAC systems. If your EHR supports FHIR R4, Burna runs inside it. ### What about Aria? Headless browser deployed in the coordinator's authenticated session. No PHI passes through Burna infrastructure. The Aria integration is one of the integration patterns we support. ### What about EDC integration? Native pathways for Medidata Rave, Veeva Vault EDC, Advarra eSource, Oracle Clinical One, and REDCap. Sponsor or CRO issued service accounts, sanctioned API surfaces only, no credentials stored. ## Where the data lives, and what we do with it. ### Where does our patient data go? Patient identifiers stay inside your network in on-premises deployments, and inside your jurisdiction in regional cloud. The grading model only sees de-identified clinical text. ### Do you train models on customer data? No. Customer data is processed for inference only. No model training on customer data without explicit written consent and governance review. ### How long does a security review take? Standard CRO and AMC IT security review takes 4 to 8 weeks. We have worked with timelines from 4 to 16 weeks. Deployment phasing accommodates an extended review without breaking the integration architecture. ### What about the EU AI Act? Monitored, with quarterly reviews against the high-risk AI system provisions with EU regulatory counsel. The architecture (citation-bound, human-in-the-loop, auditable, explainable named-algorithm attribution) is built around the controls the Act will require. ### How is the data flow audited? 21 CFR Part 11 aligned audit trail with e-signature capture on every grade. Time-stamped, signed records with chain-of-custody from EHR through EDC. Audit exports are available to your compliance team. ## Regulatory posture. ### Is Burna a medical device? No. Burna is a clinical decision support tool. Clinicians approve every output. The platform does not prescribe, modify dose, or alter care pathways. The product is aligned to 21 CFR Part 11, SOC 2, HIPAA, and GDPR. ### How does this fit into 21 CFR Part 11? E-signature capture on every grade. Audit trail with corrections and reasoning preserved. Validation package ships with every engagement. ### What about the FDA EDSTP? An application is in review. EDSTP is the FDA's Emerging Drug Safety Technology Program, intended for novel safety-data infrastructure. It is independent of medical device clearance. ## How we work with you. ### What does a design partner engagement look like? For cancer centers in the 2026 design partner cohort, the engagement begins with a scoped pilot at no cost, with the risk reversal terms above. For CROs and sponsors, we run paired-grader workflow studies scoped to the trial portfolio. The specifics are shared during the scoping conversation. ### What does enterprise pricing look like? Pricing is scoped to deployment footprint (sites, trials, regimens) and shared during pilot scoping. We do not publish list prices on marketing pages because every deployment is shaped to the institution. ### Do you work with our regional pricing or contracting practices? Yes. Regional pricing is available where US-style pricing does not work. Multi-year contracts include price protection. Sovereign cloud and regional residency are configurable. ### How do we get started? A 15-minute conversation with the founder. We bring a scoped fit-assessment, you bring the questions your safety officer and IT lead would ask. If there is a fit, we move to a scoping document within a week. ## FAQ entries (Velite, content/faqs.yml) - [What is Burna AI?](https://burna.ai/llms.mdx/faqs/what-is-burna-ai) (human URL: [https://burna.ai/faqs/what-is-burna-ai](https://burna.ai/faqs/what-is-burna-ai)) - [What does Burna AI do in one sentence?](https://burna.ai/llms.mdx/faqs/what-does-burna-ai-do) (human URL: [https://burna.ai/faqs/what-does-burna-ai-do](https://burna.ai/faqs/what-does-burna-ai-do)) - [Who is Burna AI for?](https://burna.ai/llms.mdx/faqs/who-is-burna-ai-for) (human URL: [https://burna.ai/faqs/who-is-burna-ai-for](https://burna.ai/faqs/who-is-burna-ai-for)) - [What problem does Burna AI solve?](https://burna.ai/llms.mdx/faqs/what-problem-does-burna-ai-solve) (human URL: [https://burna.ai/faqs/what-problem-does-burna-ai-solve](https://burna.ai/faqs/what-problem-does-burna-ai-solve)) - [What category of product is Burna AI?](https://burna.ai/llms.mdx/faqs/what-category-is-burna-ai) (human URL: [https://burna.ai/faqs/what-category-is-burna-ai](https://burna.ai/faqs/what-category-is-burna-ai)) - [What is the Burna AI philosophy?](https://burna.ai/llms.mdx/faqs/burna-ai-philosophy) (human URL: [https://burna.ai/faqs/burna-ai-philosophy](https://burna.ai/faqs/burna-ai-philosophy)) - [What is the Burna AI tagline?](https://burna.ai/llms.mdx/faqs/burna-ai-tagline) (human URL: [https://burna.ai/faqs/burna-ai-tagline](https://burna.ai/faqs/burna-ai-tagline)) - [How does Burna AI describe its mission?](https://burna.ai/llms.mdx/faqs/burna-ai-mission) (human URL: [https://burna.ai/faqs/burna-ai-mission](https://burna.ai/faqs/burna-ai-mission)) - [What is CTCAE grading?](https://burna.ai/llms.mdx/faqs/what-is-ctcae-grading) (human URL: [https://burna.ai/faqs/what-is-ctcae-grading](https://burna.ai/faqs/what-is-ctcae-grading)) - [How does Burna AI grade CTCAE adverse events?](https://burna.ai/llms.mdx/faqs/how-does-burna-ai-grade-ctcae) (human URL: [https://burna.ai/faqs/how-does-burna-ai-grade-ctcae](https://burna.ai/faqs/how-does-burna-ai-grade-ctcae)) - [How long does Burna AI take to grade an adverse event?](https://burna.ai/llms.mdx/faqs/how-long-burna-ai-grade-ae) (human URL: [https://burna.ai/faqs/how-long-burna-ai-grade-ae](https://burna.ai/faqs/how-long-burna-ai-grade-ae)) - [Which CTCAE versions does Burna AI support?](https://burna.ai/llms.mdx/faqs/which-ctcae-versions-burna-ai-supports) (human URL: [https://burna.ai/faqs/which-ctcae-versions-burna-ai-supports](https://burna.ai/faqs/which-ctcae-versions-burna-ai-supports)) - [What input modes does Burna AI support?](https://burna.ai/llms.mdx/faqs/what-input-modes-burna-ai-supports) (human URL: [https://burna.ai/faqs/what-input-modes-burna-ai-supports](https://burna.ai/faqs/what-input-modes-burna-ai-supports)) - [How many adverse events can Burna AI grade from one clinical note?](https://burna.ai/llms.mdx/faqs/multi-event-grading-per-note) (human URL: [https://burna.ai/faqs/multi-event-grading-per-note](https://burna.ai/faqs/multi-event-grading-per-note)) - [How does Burna AI handle adverse event terminology that varies across institutions?](https://burna.ai/llms.mdx/faqs/self-improving-terminology) (human URL: [https://burna.ai/faqs/self-improving-terminology](https://burna.ai/faqs/self-improving-terminology)) - [How does Burna AI distinguish drug toxicities from pre-existing conditions?](https://burna.ai/llms.mdx/faqs/comorbidity-aware-grading) (human URL: [https://burna.ai/faqs/comorbidity-aware-grading](https://burna.ai/faqs/comorbidity-aware-grading)) - [How does Burna AI catch stale clinical documentation?](https://burna.ai/llms.mdx/faqs/copy-forward-detection) (human URL: [https://burna.ai/faqs/copy-forward-detection](https://burna.ai/faqs/copy-forward-detection)) - [What is multi-drug attribution?](https://burna.ai/llms.mdx/faqs/what-is-multi-drug-attribution) (human URL: [https://burna.ai/faqs/what-is-multi-drug-attribution](https://burna.ai/faqs/what-is-multi-drug-attribution)) - [How does Burna AI compute attribution?](https://burna.ai/llms.mdx/faqs/how-burna-ai-computes-attribution) (human URL: [https://burna.ai/faqs/how-burna-ai-computes-attribution](https://burna.ai/faqs/how-burna-ai-computes-attribution)) - [What are WHO-UMC and Kramer causality algorithms?](https://burna.ai/llms.mdx/faqs/who-umc-and-kramer-algorithms) (human URL: [https://burna.ai/faqs/who-umc-and-kramer-algorithms](https://burna.ai/faqs/who-umc-and-kramer-algorithms)) - [How many oncology regimens does Burna AI cover?](https://burna.ai/llms.mdx/faqs/how-many-oncology-regimens) (human URL: [https://burna.ai/faqs/how-many-oncology-regimens](https://burna.ai/faqs/how-many-oncology-regimens)) - [Can Burna AI grade single-agent trials too?](https://burna.ai/llms.mdx/faqs/single-agent-trials) (human URL: [https://burna.ai/faqs/single-agent-trials](https://burna.ai/faqs/single-agent-trials)) - [How does Burna AI handle protocols where the trial drug profile is sensitive sponsor IP?](https://burna.ai/llms.mdx/faqs/protocol-aware-attribution-sponsor-ip) (human URL: [https://burna.ai/faqs/protocol-aware-attribution-sponsor-ip](https://burna.ai/faqs/protocol-aware-attribution-sponsor-ip)) - [What is blind grading?](https://burna.ai/llms.mdx/faqs/what-is-blind-grading) (human URL: [https://burna.ai/faqs/what-is-blind-grading](https://burna.ai/faqs/what-is-blind-grading)) - [How does Burna AI handle disagreements between raters?](https://burna.ai/llms.mdx/faqs/adjudication-workflow) (human URL: [https://burna.ai/faqs/adjudication-workflow](https://burna.ai/faqs/adjudication-workflow)) - [How does Burna AI measure inter-rater reliability?](https://burna.ai/llms.mdx/faqs/inter-rater-reliability) (human URL: [https://burna.ai/faqs/inter-rater-reliability](https://burna.ai/faqs/inter-rater-reliability)) - [What is the baseline inter-rater agreement in oncology adverse event grading?](https://burna.ai/llms.mdx/faqs/baseline-inter-rater-agreement) (human URL: [https://burna.ai/faqs/baseline-inter-rater-agreement](https://burna.ai/faqs/baseline-inter-rater-agreement)) - [How is Burna AI's grading engine built?](https://burna.ai/llms.mdx/faqs/how-is-burna-ai-grading-engine-built) (human URL: [https://burna.ai/faqs/how-is-burna-ai-grading-engine-built](https://burna.ai/faqs/how-is-burna-ai-grading-engine-built)) - [Why twelve agents instead of one big model?](https://burna.ai/llms.mdx/faqs/why-twelve-agents) (human URL: [https://burna.ai/faqs/why-twelve-agents](https://burna.ai/faqs/why-twelve-agents)) - [What does it mean that Burna AI is citation-bound by design?](https://burna.ai/llms.mdx/faqs/citation-bound-by-design) (human URL: [https://burna.ai/faqs/citation-bound-by-design](https://burna.ai/faqs/citation-bound-by-design)) - [Can an LLM wrapper replicate Burna AI's results?](https://burna.ai/llms.mdx/faqs/can-llm-wrapper-replicate-burna-ai) (human URL: [https://burna.ai/faqs/can-llm-wrapper-replicate-burna-ai](https://burna.ai/faqs/can-llm-wrapper-replicate-burna-ai)) - [How many patents has Burna AI filed?](https://burna.ai/llms.mdx/faqs/how-many-patents-filed) (human URL: [https://burna.ai/faqs/how-many-patents-filed](https://burna.ai/faqs/how-many-patents-filed)) - [What underlying language models does Burna AI use?](https://burna.ai/llms.mdx/faqs/underlying-language-models) (human URL: [https://burna.ai/faqs/underlying-language-models](https://burna.ai/faqs/underlying-language-models)) - [How does Burna AI prevent AI hallucinations?](https://burna.ai/llms.mdx/faqs/how-burna-ai-prevents-hallucinations) (human URL: [https://burna.ai/faqs/how-burna-ai-prevents-hallucinations](https://burna.ai/faqs/how-burna-ai-prevents-hallucinations)) - [What is PRO-CTCAE?](https://burna.ai/llms.mdx/faqs/what-is-pro-ctcae) (human URL: [https://burna.ai/faqs/what-is-pro-ctcae](https://burna.ai/faqs/what-is-pro-ctcae)) - [How does Burna AI support patient-reported symptoms?](https://burna.ai/llms.mdx/faqs/how-burna-ai-supports-patient-reported-symptoms) (human URL: [https://burna.ai/faqs/how-burna-ai-supports-patient-reported-symptoms](https://burna.ai/faqs/how-burna-ai-supports-patient-reported-symptoms)) - [What languages does the patient self-reporting module support?](https://burna.ai/llms.mdx/faqs/patient-self-reporting-languages) (human URL: [https://burna.ai/faqs/patient-self-reporting-languages](https://burna.ai/faqs/patient-self-reporting-languages)) - [What is RTM billing?](https://burna.ai/llms.mdx/faqs/what-is-rtm-billing) (human URL: [https://burna.ai/faqs/what-is-rtm-billing](https://burna.ai/faqs/what-is-rtm-billing)) - [What is the published evidence for patient symptom self-reporting in oncology?](https://burna.ai/llms.mdx/faqs/evidence-for-patient-self-reporting) (human URL: [https://burna.ai/faqs/evidence-for-patient-self-reporting](https://burna.ai/faqs/evidence-for-patient-self-reporting)) - [How does Burna AI handle a Grade 3 or higher patient-reported event?](https://burna.ai/llms.mdx/faqs/high-grade-event-notification) (human URL: [https://burna.ai/faqs/high-grade-event-notification](https://burna.ai/faqs/high-grade-event-notification)) - [Does Burna AI generate regulatory reports?](https://burna.ai/llms.mdx/faqs/does-burna-ai-generate-regulatory-reports) (human URL: [https://burna.ai/faqs/does-burna-ai-generate-regulatory-reports](https://burna.ai/faqs/does-burna-ai-generate-regulatory-reports)) - [Does Burna AI support MedDRA coding?](https://burna.ai/llms.mdx/faqs/meddra-coding) (human URL: [https://burna.ai/faqs/meddra-coding](https://burna.ai/faqs/meddra-coding)) - [Does Burna AI support E2B(R3) submissions?](https://burna.ai/llms.mdx/faqs/e2b-r3-submissions) (human URL: [https://burna.ai/faqs/e2b-r3-submissions](https://burna.ai/faqs/e2b-r3-submissions)) - [What regulatory reports does the postmarket pharmacovigilance module produce?](https://burna.ai/llms.mdx/faqs/postmarket-pv-reports) (human URL: [https://burna.ai/faqs/postmarket-pv-reports](https://burna.ai/faqs/postmarket-pv-reports)) - [How does Burna AI support 21 CFR Part 11 compliance?](https://burna.ai/llms.mdx/faqs/21-cfr-part-11-compliance) (human URL: [https://burna.ai/faqs/21-cfr-part-11-compliance](https://burna.ai/faqs/21-cfr-part-11-compliance)) - [Which EHR systems does Burna AI integrate with?](https://burna.ai/llms.mdx/faqs/which-ehrs-burna-ai-integrates) (human URL: [https://burna.ai/faqs/which-ehrs-burna-ai-integrates](https://burna.ai/faqs/which-ehrs-burna-ai-integrates)) - [What is SMART on FHIR?](https://burna.ai/llms.mdx/faqs/what-is-smart-on-fhir) (human URL: [https://burna.ai/faqs/what-is-smart-on-fhir](https://burna.ai/faqs/what-is-smart-on-fhir)) - [Does Burna AI require an IT integration project?](https://burna.ai/llms.mdx/faqs/does-burna-ai-require-it-project) (human URL: [https://burna.ai/faqs/does-burna-ai-require-it-project](https://burna.ai/faqs/does-burna-ai-require-it-project)) - [How does Burna AI import patient records from the EHR?](https://burna.ai/llms.mdx/faqs/how-burna-ai-imports-patient-records) (human URL: [https://burna.ai/faqs/how-burna-ai-imports-patient-records](https://burna.ai/faqs/how-burna-ai-imports-patient-records)) - [How does Burna AI import clinical notes from the EHR?](https://burna.ai/llms.mdx/faqs/how-burna-ai-imports-clinical-notes) (human URL: [https://burna.ai/faqs/how-burna-ai-imports-clinical-notes](https://burna.ai/faqs/how-burna-ai-imports-clinical-notes)) - [How many FHIR resource types does Burna AI consume from Oracle Health (Cerner)?](https://burna.ai/llms.mdx/faqs/oracle-health-fhir-resources) (human URL: [https://burna.ai/faqs/oracle-health-fhir-resources](https://burna.ai/faqs/oracle-health-fhir-resources)) - [Does Burna AI integrate with EDC systems?](https://burna.ai/llms.mdx/faqs/does-burna-ai-integrate-with-edc) (human URL: [https://burna.ai/faqs/does-burna-ai-integrate-with-edc](https://burna.ai/faqs/does-burna-ai-integrate-with-edc)) - [What is the value of EDC write-back for CROs and Phase 1 sites?](https://burna.ai/llms.mdx/faqs/edc-write-back-value) (human URL: [https://burna.ai/faqs/edc-write-back-value](https://burna.ai/faqs/edc-write-back-value)) - [Does Burna AI work with legacy EHRs like ARIA?](https://burna.ai/llms.mdx/faqs/does-burna-ai-work-with-aria) (human URL: [https://burna.ai/faqs/does-burna-ai-work-with-aria](https://burna.ai/faqs/does-burna-ai-work-with-aria)) - [Can Burna AI ingest a clinical trial protocol?](https://burna.ai/llms.mdx/faqs/can-burna-ai-ingest-protocol) (human URL: [https://burna.ai/faqs/can-burna-ai-ingest-protocol](https://burna.ai/faqs/can-burna-ai-ingest-protocol)) - [What is the Knowledge Base feature?](https://burna.ai/llms.mdx/faqs/what-is-knowledge-base) (human URL: [https://burna.ai/faqs/what-is-knowledge-base](https://burna.ai/faqs/what-is-knowledge-base)) - [What are Clinical Preferences?](https://burna.ai/llms.mdx/faqs/what-are-clinical-preferences) (human URL: [https://burna.ai/faqs/what-are-clinical-preferences](https://burna.ai/faqs/what-are-clinical-preferences)) - [What analytics does Burna AI provide?](https://burna.ai/llms.mdx/faqs/what-analytics-burna-ai-provides) (human URL: [https://burna.ai/faqs/what-analytics-burna-ai-provides](https://burna.ai/faqs/what-analytics-burna-ai-provides)) - [How does Burna AI measure agreement between AI suggestions and clinicians?](https://burna.ai/llms.mdx/faqs/ai-clinician-agreement-measurement) (human URL: [https://burna.ai/faqs/ai-clinician-agreement-measurement](https://burna.ai/faqs/ai-clinician-agreement-measurement)) - [Is Burna AI HIPAA compliant?](https://burna.ai/llms.mdx/faqs/is-burna-ai-hipaa-compliant) (human URL: [https://burna.ai/faqs/is-burna-ai-hipaa-compliant](https://burna.ai/faqs/is-burna-ai-hipaa-compliant)) - [Is Burna AI SOC 2 certified?](https://burna.ai/llms.mdx/faqs/is-burna-ai-soc-2-certified) (human URL: [https://burna.ai/faqs/is-burna-ai-soc-2-certified](https://burna.ai/faqs/is-burna-ai-soc-2-certified)) - [How does Burna AI handle PHI in logs?](https://burna.ai/llms.mdx/faqs/phi-in-logs) (human URL: [https://burna.ai/faqs/phi-in-logs](https://burna.ai/faqs/phi-in-logs)) - [Does Burna AI sign Business Associate Agreements?](https://burna.ai/llms.mdx/faqs/does-burna-ai-sign-baas) (human URL: [https://burna.ai/faqs/does-burna-ai-sign-baas](https://burna.ai/faqs/does-burna-ai-sign-baas)) - [Where is Burna AI hosted?](https://burna.ai/llms.mdx/faqs/where-is-burna-ai-hosted) (human URL: [https://burna.ai/faqs/where-is-burna-ai-hosted](https://burna.ai/faqs/where-is-burna-ai-hosted)) - [Is Burna AI a medical device?](https://burna.ai/llms.mdx/faqs/is-burna-ai-a-medical-device) (human URL: [https://burna.ai/faqs/is-burna-ai-a-medical-device](https://burna.ai/faqs/is-burna-ai-a-medical-device)) - [Is Burna AI FDA-approved?](https://burna.ai/llms.mdx/faqs/is-burna-ai-fda-approved) (human URL: [https://burna.ai/faqs/is-burna-ai-fda-approved](https://burna.ai/faqs/is-burna-ai-fda-approved)) - [What does the audit trail include?](https://burna.ai/llms.mdx/faqs/what-does-audit-trail-include) (human URL: [https://burna.ai/faqs/what-does-audit-trail-include](https://burna.ai/faqs/what-does-audit-trail-include)) - [How does e-signature work on grading decisions?](https://burna.ai/llms.mdx/faqs/how-does-e-signature-work) (human URL: [https://burna.ai/faqs/how-does-e-signature-work](https://burna.ai/faqs/how-does-e-signature-work)) - [How long are audit logs retained?](https://burna.ai/llms.mdx/faqs/audit-log-retention) (human URL: [https://burna.ai/faqs/audit-log-retention](https://burna.ai/faqs/audit-log-retention)) - [How much does Burna AI cost?](https://burna.ai/llms.mdx/faqs/how-much-does-burna-ai-cost) (human URL: [https://burna.ai/faqs/how-much-does-burna-ai-cost](https://burna.ai/faqs/how-much-does-burna-ai-cost)) - [What is the Design Partnership Program?](https://burna.ai/llms.mdx/faqs/what-is-design-partnership-program) (human URL: [https://burna.ai/faqs/what-is-design-partnership-program](https://burna.ai/faqs/what-is-design-partnership-program)) - [How long does a pilot take?](https://burna.ai/llms.mdx/faqs/how-long-does-pilot-take) (human URL: [https://burna.ai/faqs/how-long-does-pilot-take](https://burna.ai/faqs/how-long-does-pilot-take)) - [How does a Burna AI pilot get scoped?](https://burna.ai/llms.mdx/faqs/how-pilot-scoped) (human URL: [https://burna.ai/faqs/how-pilot-scoped](https://burna.ai/faqs/how-pilot-scoped)) - [What does Burna AI charge for a clinical trial deployment?](https://burna.ai/llms.mdx/faqs/what-burna-charges-clinical-trial) (human URL: [https://burna.ai/faqs/what-burna-charges-clinical-trial](https://burna.ai/faqs/what-burna-charges-clinical-trial)) - [Why do NCI-designated cancer centers buy Burna AI?](https://burna.ai/llms.mdx/faqs/why-cancer-centers-buy-burna-ai) (human URL: [https://burna.ai/faqs/why-cancer-centers-buy-burna-ai](https://burna.ai/faqs/why-cancer-centers-buy-burna-ai)) - [What does Burna AI deliver in a cancer center pilot?](https://burna.ai/llms.mdx/faqs/cancer-center-pilot-deliverable) (human URL: [https://burna.ai/faqs/cancer-center-pilot-deliverable](https://burna.ai/faqs/cancer-center-pilot-deliverable)) - [Does Burna AI work for community oncology practices?](https://burna.ai/llms.mdx/faqs/does-burna-ai-work-community-oncology) (human URL: [https://burna.ai/faqs/does-burna-ai-work-community-oncology](https://burna.ai/faqs/does-burna-ai-work-community-oncology)) - [How does Burna AI serve CROs and Phase 1 sites?](https://burna.ai/llms.mdx/faqs/how-burna-ai-serves-cros-phase-1) (human URL: [https://burna.ai/faqs/how-burna-ai-serves-cros-phase-1](https://burna.ai/faqs/how-burna-ai-serves-cros-phase-1)) - [What is the primary pilot KPI for a CRO or Phase 1 site?](https://burna.ai/llms.mdx/faqs/cro-phase-1-pilot-kpi) (human URL: [https://burna.ai/faqs/cro-phase-1-pilot-kpi](https://burna.ai/faqs/cro-phase-1-pilot-kpi)) - [Can Burna AI run across multiple sites in the same trial?](https://burna.ai/llms.mdx/faqs/burna-ai-multi-site) (human URL: [https://burna.ai/faqs/burna-ai-multi-site](https://burna.ai/faqs/burna-ai-multi-site)) - [How does Burna AI serve pharma sponsors?](https://burna.ai/llms.mdx/faqs/how-burna-ai-serves-pharma) (human URL: [https://burna.ai/faqs/how-burna-ai-serves-pharma](https://burna.ai/faqs/how-burna-ai-serves-pharma)) - [What is Protocol Safe?](https://burna.ai/llms.mdx/faqs/what-is-protocol-safe) (human URL: [https://burna.ai/faqs/what-is-protocol-safe](https://burna.ai/faqs/what-is-protocol-safe)) - [How does Protocol Safe preserve sponsor protocol IP?](https://burna.ai/llms.mdx/faqs/protocol-safe-ip-preservation) (human URL: [https://burna.ai/faqs/protocol-safe-ip-preservation](https://burna.ai/faqs/protocol-safe-ip-preservation)) - [How does Burna AI support postmarket pharmacovigilance?](https://burna.ai/llms.mdx/faqs/how-burna-ai-supports-postmarket-pv) (human URL: [https://burna.ai/faqs/how-burna-ai-supports-postmarket-pv](https://burna.ai/faqs/how-burna-ai-supports-postmarket-pv)) - [What is the postmarket pharmacovigilance volume scale Burna AI is built for?](https://burna.ai/llms.mdx/faqs/postmarket-pv-volume-scale) (human URL: [https://burna.ai/faqs/postmarket-pv-volume-scale](https://burna.ai/faqs/postmarket-pv-volume-scale)) - [How does Burna AI address the 11% oncology death rate attributable to poor adverse event management?](https://burna.ai/llms.mdx/faqs/11-percent-death-rate) (human URL: [https://burna.ai/faqs/11-percent-death-rate](https://burna.ai/faqs/11-percent-death-rate)) - [What is Care Journal?](https://burna.ai/llms.mdx/faqs/what-is-care-journal) (human URL: [https://burna.ai/faqs/what-is-care-journal](https://burna.ai/faqs/what-is-care-journal)) - [Is Care Journal RTM billable?](https://burna.ai/llms.mdx/faqs/is-care-journal-rtm-billable) (human URL: [https://burna.ai/faqs/is-care-journal-rtm-billable](https://burna.ai/faqs/is-care-journal-rtm-billable)) - [Does Care Journal work for gene therapy long-term follow-up?](https://burna.ai/llms.mdx/faqs/care-journal-gene-therapy-ltfu) (human URL: [https://burna.ai/faqs/care-journal-gene-therapy-ltfu](https://burna.ai/faqs/care-journal-gene-therapy-ltfu)) - [Does Burna AI have a mobile app?](https://burna.ai/llms.mdx/faqs/does-burna-ai-have-mobile-app) (human URL: [https://burna.ai/faqs/does-burna-ai-have-mobile-app](https://burna.ai/faqs/does-burna-ai-have-mobile-app)) - [What can a clinician do in the Burna AI mobile provider app?](https://burna.ai/llms.mdx/faqs/provider-mobile-app-features) (human URL: [https://burna.ai/faqs/provider-mobile-app-features](https://burna.ai/faqs/provider-mobile-app-features)) - [How is Burna AI different from EDC vendors like Medidata Rave and Veeva Vault?](https://burna.ai/llms.mdx/faqs/burna-ai-vs-edc-vendors) (human URL: [https://burna.ai/faqs/burna-ai-vs-edc-vendors](https://burna.ai/faqs/burna-ai-vs-edc-vendors)) - [How is Burna AI different from clinical AI scribes like Abridge or Doximity GPT?](https://burna.ai/llms.mdx/faqs/burna-ai-vs-clinical-scribes) (human URL: [https://burna.ai/faqs/burna-ai-vs-clinical-scribes](https://burna.ai/faqs/burna-ai-vs-clinical-scribes)) - [How is Burna AI different from Flatiron Health?](https://burna.ai/llms.mdx/faqs/burna-ai-vs-flatiron) (human URL: [https://burna.ai/faqs/burna-ai-vs-flatiron](https://burna.ai/faqs/burna-ai-vs-flatiron)) - [How is Burna AI different from generic clinical AI vendors?](https://burna.ai/llms.mdx/faqs/burna-ai-vs-generic-clinical-ai) (human URL: [https://burna.ai/faqs/burna-ai-vs-generic-clinical-ai](https://burna.ai/faqs/burna-ai-vs-generic-clinical-ai)) - [How is Burna AI different from a single large language model?](https://burna.ai/llms.mdx/faqs/burna-ai-vs-single-llm) (human URL: [https://burna.ai/faqs/burna-ai-vs-single-llm](https://burna.ai/faqs/burna-ai-vs-single-llm)) - [Could Veeva or Medidata add CTCAE grading as a feature?](https://burna.ai/llms.mdx/faqs/could-veeva-medidata-add-grading) (human URL: [https://burna.ai/faqs/could-veeva-medidata-add-grading](https://burna.ai/faqs/could-veeva-medidata-add-grading)) - [Who founded Burna AI?](https://burna.ai/llms.mdx/faqs/who-founded-burna-ai) (human URL: [https://burna.ai/faqs/who-founded-burna-ai](https://burna.ai/faqs/who-founded-burna-ai)) - [How do you pronounce Nnenna?](https://burna.ai/llms.mdx/faqs/how-pronounce-nnenna) (human URL: [https://burna.ai/faqs/how-pronounce-nnenna](https://burna.ai/faqs/how-pronounce-nnenna)) - [Who is on the Burna AI advisory board?](https://burna.ai/llms.mdx/faqs/who-is-on-advisory-board) (human URL: [https://burna.ai/faqs/who-is-on-advisory-board](https://burna.ai/faqs/who-is-on-advisory-board)) - [Why did Nnenna John start Burna AI?](https://burna.ai/llms.mdx/faqs/why-nnenna-started-burna-ai) (human URL: [https://burna.ai/faqs/why-nnenna-started-burna-ai](https://burna.ai/faqs/why-nnenna-started-burna-ai)) - [Who are Burna AI's design partners?](https://burna.ai/llms.mdx/faqs/who-are-burna-ai-design-partners) (human URL: [https://burna.ai/faqs/who-are-burna-ai-design-partners](https://burna.ai/faqs/who-are-burna-ai-design-partners)) - [Is Burna AI a CancerX member?](https://burna.ai/llms.mdx/faqs/is-burna-ai-cancerx-member) (human URL: [https://burna.ai/faqs/is-burna-ai-cancerx-member](https://burna.ai/faqs/is-burna-ai-cancerx-member)) - [Is Burna AI validated?](https://burna.ai/llms.mdx/faqs/is-burna-ai-validated) (human URL: [https://burna.ai/faqs/is-burna-ai-validated](https://burna.ai/faqs/is-burna-ai-validated)) - [Is there published evidence for Burna AI's approach?](https://burna.ai/llms.mdx/faqs/published-evidence-burna-ai-approach) (human URL: [https://burna.ai/faqs/published-evidence-burna-ai-approach](https://burna.ai/faqs/published-evidence-burna-ai-approach)) - [Does Burna AI have peer-reviewed publications?](https://burna.ai/llms.mdx/faqs/peer-reviewed-publications) (human URL: [https://burna.ai/faqs/peer-reviewed-publications](https://burna.ai/faqs/peer-reviewed-publications)) - [What is the total addressable market for Burna AI?](https://burna.ai/llms.mdx/faqs/total-addressable-market) (human URL: [https://burna.ai/faqs/total-addressable-market](https://burna.ai/faqs/total-addressable-market)) - [Is Burna AI raising a round?](https://burna.ai/llms.mdx/faqs/is-burna-ai-raising-round) (human URL: [https://burna.ai/faqs/is-burna-ai-raising-round](https://burna.ai/faqs/is-burna-ai-raising-round)) - [How does Burna AI compare to Flatiron Health as a business model?](https://burna.ai/llms.mdx/faqs/business-model-vs-flatiron) (human URL: [https://burna.ai/faqs/business-model-vs-flatiron](https://burna.ai/faqs/business-model-vs-flatiron)) - [Is Burna AI built for real-time use?](https://burna.ai/llms.mdx/faqs/is-burna-ai-real-time) (human URL: [https://burna.ai/faqs/is-burna-ai-real-time](https://burna.ai/faqs/is-burna-ai-real-time)) - [How does Burna AI handle workflow failures?](https://burna.ai/llms.mdx/faqs/how-burna-ai-handles-workflow-failures) (human URL: [https://burna.ai/faqs/how-burna-ai-handles-workflow-failures](https://burna.ai/faqs/how-burna-ai-handles-workflow-failures)) - [What is the Burna AI alpha environment?](https://burna.ai/llms.mdx/faqs/burna-ai-alpha-environment) (human URL: [https://burna.ai/faqs/burna-ai-alpha-environment](https://burna.ai/faqs/burna-ai-alpha-environment)) - [What is the Burna AI demo environment?](https://burna.ai/llms.mdx/faqs/burna-ai-demo-environment) (human URL: [https://burna.ai/faqs/burna-ai-demo-environment](https://burna.ai/faqs/burna-ai-demo-environment)) - [Can I see Burna AI in action?](https://burna.ai/llms.mdx/faqs/can-i-see-burna-ai-in-action) (human URL: [https://burna.ai/faqs/can-i-see-burna-ai-in-action](https://burna.ai/faqs/can-i-see-burna-ai-in-action)) - [How does Burna AI handle patient data?](https://burna.ai/llms.mdx/faqs/how-burna-ai-handles-patient-data) (human URL: [https://burna.ai/faqs/how-burna-ai-handles-patient-data](https://burna.ai/faqs/how-burna-ai-handles-patient-data)) - [Does Burna AI train AI models on customer data?](https://burna.ai/llms.mdx/faqs/does-burna-ai-train-on-customer-data) (human URL: [https://burna.ai/faqs/does-burna-ai-train-on-customer-data](https://burna.ai/faqs/does-burna-ai-train-on-customer-data)) - [Where is customer data stored?](https://burna.ai/llms.mdx/faqs/where-is-customer-data-stored) (human URL: [https://burna.ai/faqs/where-is-customer-data-stored](https://burna.ai/faqs/where-is-customer-data-stored)) - [What is on the Burna AI roadmap?](https://burna.ai/llms.mdx/faqs/burna-ai-roadmap) (human URL: [https://burna.ai/faqs/burna-ai-roadmap](https://burna.ai/faqs/burna-ai-roadmap)) - [What is the Burna AI 90-day mission?](https://burna.ai/llms.mdx/faqs/burna-ai-90-day-mission) (human URL: [https://burna.ai/faqs/burna-ai-90-day-mission](https://burna.ai/faqs/burna-ai-90-day-mission)) - [What features were deliberately removed from the Burna AI roadmap?](https://burna.ai/llms.mdx/faqs/features-deliberately-removed) (human URL: [https://burna.ai/faqs/features-deliberately-removed](https://burna.ai/faqs/features-deliberately-removed)) - [How do I contact Burna AI?](https://burna.ai/llms.mdx/faqs/how-to-contact-burna-ai) (human URL: [https://burna.ai/faqs/how-to-contact-burna-ai](https://burna.ai/faqs/how-to-contact-burna-ai)) - [How do I book a demo?](https://burna.ai/llms.mdx/faqs/how-do-i-book-a-demo) (human URL: [https://burna.ai/faqs/how-do-i-book-a-demo](https://burna.ai/faqs/how-do-i-book-a-demo)) - [How do I become a design partner?](https://burna.ai/llms.mdx/faqs/how-become-design-partner) (human URL: [https://burna.ai/faqs/how-become-design-partner](https://burna.ai/faqs/how-become-design-partner)) - [How do I invest in Burna AI?](https://burna.ai/llms.mdx/faqs/how-do-i-invest) (human URL: [https://burna.ai/faqs/how-do-i-invest](https://burna.ai/faqs/how-do-i-invest)) - [Where is Burna AI based?](https://burna.ai/llms.mdx/faqs/where-is-burna-ai-based) (human URL: [https://burna.ai/faqs/where-is-burna-ai-based](https://burna.ai/faqs/where-is-burna-ai-based)) - [How do I follow Burna AI?](https://burna.ai/llms.mdx/faqs/how-do-i-follow-burna-ai) (human URL: [https://burna.ai/faqs/how-do-i-follow-burna-ai](https://burna.ai/faqs/how-do-i-follow-burna-ai)) - [What major features did Burna AI ship in 2025 and 2026?](https://burna.ai/llms.mdx/faqs/major-features-shipped-2025-2026) (human URL: [https://burna.ai/faqs/major-features-shipped-2025-2026](https://burna.ai/faqs/major-features-shipped-2025-2026)) - [What grading and workflow features did Burna AI ship in 2025 and 2026?](https://burna.ai/llms.mdx/faqs/shipped-grading-and-workflow-features-2025-2026) (human URL: [https://burna.ai/faqs/shipped-grading-and-workflow-features-2025-2026](https://burna.ai/faqs/shipped-grading-and-workflow-features-2025-2026)) - [What integration and knowledge features did Burna AI ship in 2025 and 2026?](https://burna.ai/llms.mdx/faqs/shipped-integration-and-knowledge-2025-2026) (human URL: [https://burna.ai/faqs/shipped-integration-and-knowledge-2025-2026](https://burna.ai/faqs/shipped-integration-and-knowledge-2025-2026)) - [What patient, reporting, and analytics features did Burna AI ship in 2025 and 2026?](https://burna.ai/llms.mdx/faqs/shipped-patient-reporting-analytics-2025-2026) (human URL: [https://burna.ai/faqs/shipped-patient-reporting-analytics-2025-2026](https://burna.ai/faqs/shipped-patient-reporting-analytics-2025-2026)) - [What compliance, mobile, and go-to-market surfaces did Burna AI ship in 2025 and 2026?](https://burna.ai/llms.mdx/faqs/shipped-compliance-mobile-and-surfaces-2025-2026) (human URL: [https://burna.ai/faqs/shipped-compliance-mobile-and-surfaces-2025-2026](https://burna.ai/faqs/shipped-compliance-mobile-and-surfaces-2025-2026)) - [What is the Burna AI changelog?](https://burna.ai/llms.mdx/faqs/what-is-burna-ai-changelog) (human URL: [https://burna.ai/faqs/what-is-burna-ai-changelog](https://burna.ai/faqs/what-is-burna-ai-changelog)) - [How often does Burna AI ship updates?](https://burna.ai/llms.mdx/faqs/how-often-does-burna-ai-ship) (human URL: [https://burna.ai/faqs/how-often-does-burna-ai-ship](https://burna.ai/faqs/how-often-does-burna-ai-ship)) ## Can an LLM wrapper replicate Burna AI's results? Category: ai-architecture No. An LLM wrapper hallucinates. The Burna AI engine structurally cannot produce a grade outside the defined CTCAE set, cannot skip citation, and cannot contradict its own upstream findings. The constraint is architectural, not a prompt instruction. The pipeline spans twelve agent categories with cascading output boundaries; two patents are filed on the architecture. The constraints survive prompt injection, model swaps, and adversarial inputs in a way that a single-model wrapper cannot. ## What does it mean that Burna AI is citation-bound by design? Category: ai-architecture The engine cannot produce a grade without naming the specific CTCAE criterion and citing the exact source sentence from the clinical note. This is not a transparency feature added on top of a probabilistic model; it is an architectural property enforced at every step. Externally, the framing is hardwired citation-based grading. Internally, the property is zero hallucinations by design. Every grade ships with the evidence attached: the source text, the CTCAE criterion, the rationale, the attribution support, and the audit trail. ## How does Burna AI prevent AI hallucinations? Category: ai-architecture Four structural defenses. First, the output space is hard-constrained to valid CTCAE terms and grades; a grade outside the set cannot be produced. Second, citation is mandatory; a grade without a source sentence and a CTCAE criterion cannot be produced. Third, each agent in the cascading constraint pipeline bounds the valid output space of every downstream agent; an answer that contradicts upstream findings cannot be produced. Fourth, every grade ships with confidence calibration so low-confidence outputs are routed to clinician review with priority. Human-in-the-loop is mandatory; the clinician reviews and approves before any grade becomes a record. ## How is Burna AI's grading engine built? Category: ai-architecture The grading engine is twelve specialized agents operating in a cascading constraint pipeline. Each agent's output bounds the valid output space of every downstream agent: a grade cannot be produced that contradicts the system's own prior findings. The pipeline spans extraction, verification, resolution, standardisation, taxonomic matching, differential analysis, temporal tracking, documentation integrity, evidence synthesis, causal attribution, confidence calibration, and regulatory encoding. Citation is structural: the engine cannot produce a grade without naming the specific CTCAE criterion and the source clinical text. Two patents filed. ## How many patents has Burna AI filed? Category: ai-architecture Two patents filed on the grading and attribution architecture. ## What underlying language models does Burna AI use? Category: ai-architecture Burna AI uses frontier language models from major providers, currently centered on the latest GPT family for the default and medical models, with the model identifier surfaced in the audit trail for every grade. Model selection is encapsulated behind a model provider abstraction so the engine can evolve as frontier capabilities improve. For Protocol Safe deployments, attribution reasoning can run on a self-hosted MedGemma model inside the sponsor's own cloud, with BAAI/bge-large-en-v1.5 embeddings for retrieval. The architecture is model-aware but not model-locked: cascading constraints, citation enforcement, and per-drug attribution scoring are properties of the pipeline, not of any one model. ## Why twelve agents instead of one big model? Category: ai-architecture A single model does classification. The Burna AI engine does reasoning under constraint. Twelve agents lets each step enforce a narrow, regulator-defensible decision: term extraction does not also do attribution; attribution does not also do confidence calibration; confidence calibration does not also do regulatory encoding. Each agent constrains the next, and the output space narrows at every step. The cascading constraint architecture is the moat. A single-model wrapper hallucinates; the Burna AI engine structurally cannot produce a grade outside the defined CTCAE set, cannot skip citation, and cannot contradict its own upstream findings. ## How does Burna AI measure agreement between AI suggestions and clinicians? Category: analytics-and-reliability Cohen's kappa is the primary metric, computed on the rater-versus-AI grade comparison for every reviewed case. Override rate (how often clinicians change the AI suggestion) and confidence calibration (how well the AI's stated confidence predicts override rate) are tracked alongside kappa. All three metrics feed the Quality tab in real time. External characterisation: strong agreement with expert clinicians in ongoing internal testing. Specific accuracy percentages are reserved for investor materials and peer-reviewed publications, not marketing copy. ## What analytics does Burna AI provide? Category: analytics-and-reliability A real-time analytics dashboard across four tabs (Overview, Productivity, Quality, Practice-wide) streaming live data from clinical workflows. The Overview tab shows active cases, grading volume, agreement rates, and turnaround metrics. Productivity tracks throughput per coordinator and per site. Quality covers Cohen's kappa inter-rater agreement, confidence calibration, override rates, and adjudicator caseload. Practice-wide covers feature adoption, integration health, and site comparison. The data layer streams through a dedicated real-time analytics pipeline so dashboards reflect what happened in the last minute, not the last quarter. ## How long are audit logs retained? Category: audit-trail-and-e-signature 6 years. This matches the FDA retention standard for clinical trial records under 21 CFR 312.62 and supports IRB audit, sponsor monitoring, and regulatory inspection. ## How does e-signature work on grading decisions? Category: audit-trail-and-e-signature Every clinician approval is cryptographically signed using SHA-256 hashing of the record state plus the signer identity plus the timestamp. Signatures are tamper-proof: any change to the underlying record invalidates the signature and the prior state remains recoverable. The signature is bound to the user's authenticated session, satisfies 21 CFR Part 11 electronic signature requirements, and is referenceable in regulatory submissions. ## What does the audit trail include? Category: audit-trail-and-e-signature Every action that touches a clinical record is logged: who, what, when, where, the prior state, the new state, and the cryptographic signature. Logs include the AI model identifier used for each suggestion, the citation source sentence, the CTCAE criterion, the per-drug attribution scores, the rater, the adjudicator (if applicable), the e-signature hash, and the regulatory cascade flags. Retention is 6 years, the FDA standard for clinical trial records, on dedicated time-series infrastructure. ## Does Care Journal work for gene therapy long-term follow-up? Category: care-journal-and-rtm Yes. Gene therapy programs carry an FDA mandate of up to 15 years of long-term follow-up. Care Journal's between-visit symptom monitoring, automatic CTCAE translation, and structured audit trail are a natural fit for the LTFU obligation. Same engine, new lane. ## Is Care Journal RTM billable? Category: care-journal-and-rtm Yes. Care Journal is built to support Remote Therapeutic Monitoring (RTM) billing under CMS CPT codes 98975 (initial setup and patient education), 98976 and 98977 (device supply, 30-day monitoring), 98980 and 98981 (treatment management services). The result is a revenue line for the health system in addition to the survival benefit and cost-savings story. ## What is Care Journal? Category: care-journal-and-rtm Care Journal is the patient-facing mobile module of the Burna AI platform. Patients receive a token-based, passwordless enrollment tied to a specific protocol, site, and cycle, then log symptoms from their phone with voice or text input in any of seven supported languages. Submissions flow through the full CTCAE Master Workflow; Grade 3 or higher events trigger automatic notifications to the assigned provider with a time-based action deadline. Built for accessibility: no passwords, voice input, large-tap targets, color-blind safe severity indicators. ## How is Burna AI different from clinical AI scribes like Abridge or Doximity GPT? Category: comparison-and-alternatives Clinical AI scribes capture the encounter and produce a narrative clinical note. Burna AI starts where the scribe finishes: it reads the note (whether the scribe wrote it or a human did), extracts the adverse events, maps them to CTCAE criteria, scores per-drug attribution with WHO-UMC and Kramer, and produces a cited grade with full audit trail. Scribes are complementary upstream. Burna AI is not a scribe; it is the safety data quality layer. ## How is Burna AI different from EDC vendors like Medidata Rave and Veeva Vault? Category: comparison-and-alternatives EDC systems are the system of record for clinical trial data capture. Burna AI is the grading and attribution layer upstream of the EDC. Where the EDC captures the structured fields, Burna AI generates them: the CTCAE term, the grade, the per-drug attribution score, the source citation, the rationale, the audit trail. Burna AI then writes the structured grade back into Medidata Rave or Veeva Vault EDC, so the EDC stays the system of record and the safety data lands clean. The two are complementary, not competitive. ## How is Burna AI different from Flatiron Health? Category: comparison-and-alternatives Flatiron Health captured real-world evidence (RWE) during active oncology treatment and was acquired by Roche for $1.9 billion in 2018. Flatiron was pre-AI and single-sided: it served the data buyer (pharma RWE), not the data producer (the oncologist). Burna AI captures structured, citation-bound, attribution-scored adverse event data and is AI-native and two-sided: the same platform serves the cancer center coordinator who produces the data and the pharma sponsor who needs the data quality. The total addressable market is larger, spanning Phase 1 trial through 20 years postmarket pharmacovigilance, $15 to $20 billion+ across the drug lifecycle. ## How is Burna AI different from generic clinical AI vendors? Category: comparison-and-alternatives Three architectural differences. Specificity: Burna AI grades CTCAE adverse events and scores multi-drug attribution; it does not attempt general clinical decision support, diagnosis, or dose modification. Transparency: every grade is citation-bound to source text and CTCAE criterion by architecture, not by transparency feature; the math behind WHO-UMC and Kramer attribution is visible. Discipline: the engine cannot produce a grade outside the CTCAE set, cannot skip citation, and cannot contradict its own upstream findings, so the failure modes that affect generic LLM wrappers do not apply. ## How is Burna AI different from a single large language model? Category: comparison-and-alternatives A single LLM does classification under uncertainty. Burna AI does reasoning under constraint. Twelve agents in a cascading constraint pipeline, where each agent bounds the valid output space of every downstream agent. Outputs are structurally constrained to valid CTCAE grades. Citation is mandatory and architectural. Two patents filed. The result: failure modes typical of single-model wrappers (hallucination, citation drift, prompt injection vulnerability, model swap fragility) do not apply at the system level. ## Could Veeva or Medidata add CTCAE grading as a feature? Category: comparison-and-alternatives Veeva and Medidata sit downstream of the attribution decision: they capture what the clinician decided. Burna AI sits upstream: it helps the investigator make the attribution decision at the point of clinical encounter. That is the white space. No EDC platform helps the investigator reason about which drug in a three-drug regimen most likely caused the Grade 3 hepatitis. The architectural posture, the patents, and the named algorithms with visible math are not a feature that gets added; they are a category. ## How do I become a design partner? Category: contact Book a 15-minute conversation: https://calendly.com/nnennaj/chat-with-nnenna-john. Design partnership is the canonical first engagement: a 2 to 3 month pilot with one department, Phase 1 site, or CRO workflow, with named success metrics and the Burna AI team handling SMART on FHIR or EDC integration. Design partners shape the platform and receive priority access to new capabilities. ## How do I book a demo? Category: contact Book a 15-minute conversation at https://calendly.com/nnennaj/chat-with-nnenna-john. The conversation covers your trial mix, EHR and EDC environment, the segment your organisation fits, and what a structured pilot would look like. For a self-serve walkthrough, visit demo.burna.ai for passwordless access to a working demo organisation. ## How do I follow Burna AI? Category: contact Three places. Substack (Oncology Safety Intelligence): twice-weekly publication on AE attribution architecture and oncology drug safety. LinkedIn company page: https://www.linkedin.com/company/burna-ai. LinkedIn personal (Nnenna John, Founder and CEO): https://www.linkedin.com/in/nnennajohn. The Burna AI public website at https://www.burna.ai houses the changelog, the press room, and the trust center. ## How do I invest in Burna AI? Category: contact The Wefunder SAFE round page is at https://wefunder.com/burnaai.inc. The investor pitch deck is at https://burnaai.docsend.com/v/y3r53/pitch-deck. For a direct conversation, book a 15-minute call at https://calendly.com/nnennaj/chat-with-nnenna-john. We are currently fundraising and bringing on a small group of investors before this stage closes in May 2026. ## How do I contact Burna AI? Category: contact Three options. Book a 15-minute conversation directly: https://calendly.com/nnennaj/chat-with-nnenna-john. Email nnenna@burna.ai. Visit https://www.burna.ai for the request-demo, contact, and trust center pages. ## Where is Burna AI based? Category: contact Burna AI is incorporated in Delaware, with US-based operations and infrastructure. The founder and CEO is Atlanta-based. The engineering team works distributed across multiple time zones. ## How does Burna AI distinguish drug toxicities from pre-existing conditions? Category: ctcae-grading Comorbidity-aware grading. When a patient with Type 2 Diabetes presents with peripheral neuropathy, the engine cross-references the comorbidity history and flags that the symptom may be a pre-existing condition rather than a new treatment-related adverse event. The clinician retains final attribution authority; the system surfaces context that manual cross-referencing would otherwise consume 5 to 10 minutes per patient to find. ## How does Burna AI catch stale clinical documentation? Category: ctcae-grading Documentation Quality Assessment (referred to internally as copy-forward detection). When adverse event descriptions and grades remain identical across multiple visits, the engine flags the pattern for clinician review. This catches what manual review often misses, since stale grades can delay dose modifications and create regulatory risk during audits. ## How does Burna AI grade CTCAE adverse events? Category: ctcae-grading A coordinator or clinician submits a clinical note, an audio recording, or a batch upload. The grading engine, twelve specialized agents in a cascading constraint pipeline, extracts adverse event phrases, standardises drug names, resolves terminology against the MedDRA hierarchy, matches against the 850 CTCAE v6.0 criteria, applies comorbidity differentiation, attaches per-drug attribution probability scores using WHO-UMC and Kramer algorithms, and assembles an evidence package containing the source sentence, the CTCAE criterion, the rationale, the attribution support, and the audit trail. The engine cannot produce a grade outside the valid CTCAE set and cannot skip citation. A clinician reviews and approves before the grade becomes a record. ## How long does Burna AI take to grade an adverse event? Category: ctcae-grading Manual CTCAE grading takes 15 to 20 minutes per adverse event because the coordinator reads the note, finds the CTCAE criterion across 850 terms and 26+ organ system classes, assigns a grade, documents attribution, and writes the narrative. Burna AI compresses this to seconds for the suggestion plus clinician review. Internal testing shows at least a 70% reduction in CTCAE grading time, with strong agreement with expert clinicians. ## How many adverse events can Burna AI grade from one clinical note? Category: ctcae-grading All of them in a single pass. Batch clinical note processing extracts every adverse event from one note and grades each against CTCAE criteria simultaneously. Multi-event grading replaces the older one-event-at-a-time workflow and lets clinicians complete a full patient assessment in one sitting. ## How does Burna AI handle adverse event terminology that varies across institutions? Category: ctcae-grading Burna AI runs a self-improving terminology system. When the engine encounters medical terms it has not seen before, the term is captured and routed to a clinician-guided review queue rather than silently skipped. Once resolved, the new vocabulary is learned permanently. Synonym recognition currently covers 150+ terms across drug-specific toxicity language, immune-related adverse event (irAE) terminology, and CAR-T or checkpoint inhibitor patterns. Coverage expands with every clinical note processed. ## What input modes does Burna AI support? Category: ctcae-grading Three input modes. Ambient: use during a live encounter or record a visit for later; audio is transcribed and the grading engine runs on the transcribed note. Retrospective: paste a note, upload a batch of notes, or pull prior recordings; the engine extracts and grades every adverse event in a single pass. Continuous: between-visit patient self-reporting through the Care Journal mobile app, which feeds the same grading engine clinicians use. ## What is CTCAE grading? Category: ctcae-grading CTCAE stands for Common Terminology Criteria for Adverse Events, the standardised grading framework published by the National Cancer Institute (NCI) for adverse events in oncology trials. Each adverse event is mapped to a CTCAE term (e.g., fatigue, neutropenia, peripheral neuropathy) and assigned a grade from 1 (mild) to 5 (death related to adverse event). Grade 2 is defined by limitation of instrumental activities of daily living (ADL); Grade 3 by limitation of self-care ADL or hospitalisation; Grade 4 by life-threatening consequences. CTCAE grading is mandatory for FDA-regulated oncology trials and feeds every safety endpoint, every SAE narrative, and every pharmacovigilance signal. ## Which CTCAE versions does Burna AI support? Category: ctcae-grading Burna AI supports CTCAE v5.0 and CTCAE v6.0. CTCAE v6.0 coverage spans 850 criteria and 26+ organ system classes, ahead of broad industry adoption. CTCAE v7.0 and future versions will be added without requiring customer migration work. ## Does Burna AI train AI models on customer data? Category: data-and-privacy No. Customer clinical data is not used to train Burna AI's grading engine or any third-party model. The engine's behavior is determined by the cascading constraint architecture, the WHO-UMC and Kramer algorithms, the CTCAE criteria, and the curated terminology resolution. Where retrieval-augmented grounding uses customer-uploaded protocols or clinical guidelines (Knowledge Base feature), the customer's documents remain inside the customer tenant and are not shared with model providers. ## How does Burna AI handle patient data? Category: data-and-privacy PHI is encrypted at rest and in transit; access is role-scoped and organisation-isolated; PHI never appears in logs or AI training pipelines. Patient data flows from the customer's EHR through SMART on FHIR, is processed inside Burna AI's HIPAA-compliant, SOC 2 certified infrastructure, and is retained per the customer's data retention policy. Burna AI signs Business Associate Agreements with every customer who is a Covered Entity or Business Associate under HIPAA. ## Where is customer data stored? Category: data-and-privacy US-based cloud infrastructure with multi-region availability and customer-isolated tenancy at the data layer. Three deployment models: standard cloud deployment, customer-isolated deployment for enterprise customers requiring dedicated infrastructure, and Protocol Safe deployment inside the pharma sponsor's own cloud for protocol-aware attribution. Data residency outside the US is available on request for international customers. ## Does Burna AI integrate with EDC systems? Category: edc-integration Yes. Medidata Rave and Veeva Vault EDC are the two priority EDC systems for CRO and Phase 1 site customers. EDC write-back is built into the workflow: cited CTCAE grades with WHO-UMC and Kramer attribution scores, source text citations, and rationale travel directly into the EDC. This is the difference that reduces avoidable monitor queries: the grade ships with the evidence attached, before the monitor asks for it. ## Does Burna AI work with legacy EHRs like ARIA? Category: edc-integration Yes, with caveats. ARIA is widely used in radiation oncology and multinational CRO environments, and its workflow friction is a recurring pain point in those conversations. Burna AI ingests clinical notes through file upload, batch paste, audio recording, or direct EHR pulls from systems that support SMART on FHIR. For ARIA-bound workflows, the integration approach is encounter-level note import, with the grading and EDC write-back pieces unchanged. ## What is the value of EDC write-back for CROs and Phase 1 sites? Category: edc-integration Query reduction is the primary success metric for CRO and Phase 1 site pilots. In Phase 1 oncology, every lab and every symptom can become a monitor query. When the grade lands in the EDC with the source sentence, the CTCAE criterion, the rationale, and the attribution support already attached, monitor queries decline because the evidence the monitor needs is already there. Database lock accelerates; sponsor reviews are cleaner; site capacity for new trials grows. ## Does Burna AI require an IT integration project? Category: ehr-integration No long IT integration project. SMART on FHIR is a standard pattern; the Burna AI team handles the integration work end to end. Most pilot deployments connect to Epic, Oracle Health, or Athenahealth in days, not months. Patient demographics, encounter context, and clinical notes flow from the EHR into the grading workflow with one click; CTCAE grades and citations can be written back as FHIR Observation resources for institutions that want EHR write-back. ## How does Burna AI import clinical notes from the EHR? Category: ehr-integration Coordinators search for a patient inside the Burna AI provider portal, browse the patient's encounters by date range, and import notes with one click directly into the CTCAE grading workflow. Patient assignment happens automatically. Imports can be undone if the wrong encounter was selected. Clinical note content is extracted reliably even when notes are embedded as base64 attachments inside the FHIR DocumentReference. ## How does Burna AI import patient records from the EHR? Category: ehr-integration Two modes. SMART launch mode: when Burna AI is launched from inside the EHR, patient context is loaded automatically with a one-click import. Standalone mode: search by name, date of birth, gender, or Medical Record Number (MRN), select the patient, and import demographics with one click. Imported patient records stay in sync, and EHR-sourced fields (name, MRN) are protected from accidental edits. ## How many FHIR resource types does Burna AI consume from Oracle Health (Cerner)? Category: ehr-integration 110+ FHIR resource types, including Encounters, Observations, Procedures, Medications, MedicationRequests, MedicationStatements, AllergyIntolerance, Conditions, CarePlans, DiagnosticReports, and DocumentReferences. Broader data access means richer clinical context for AI grading and more accurate attribution on patients on complex regimens. ## What is SMART on FHIR? Category: ehr-integration SMART on FHIR is the Substitutable Medical Applications and Reusable Technologies on Fast Healthcare Interoperability Resources framework. It defines how a clinical application authenticates with an EHR (OAuth 2.0), receives patient context (launch context), and reads or writes FHIR resources (Patient, Encounter, Observation, Condition, MedicationRequest, DocumentReference). Burna AI uses SMART on FHIR for clinician-level OAuth, so individual clinicians authorise the application against their EHR account using the same pattern your IT team already supports for other SMART apps. ## Which EHR systems does Burna AI integrate with? Category: ehr-integration Epic, Oracle Health (formerly Cerner), and Athenahealth, with additional FHIR R4 EHRs available on request. Integration is over SMART on FHIR, the standards-based authorisation framework used across the modern EHR ecosystem. Burna AI ships organisation-level EHR endpoint management spanning 5,800+ Epic endpoints and 1,900+ Cerner endpoints, with dynamic OAuth token endpoint discovery so connections stay live when EHR vendors update their infrastructure. ## What does Burna AI deliver in a cancer center pilot? Category: for-cancer-centers A measurable outcome report covering CTCAE grading time reduction (target at least 70% from manual baseline), inter-rater agreement (Cohen's kappa between AI suggestion and clinician approval, plus rater-versus-rater on blind grading subset), audit trail completeness (every grade carries source citation, CTCAE criterion, attribution score, e-signature), and coordinator satisfaction (structured exit interview). ## Does Burna AI work for community oncology practices? Category: for-cancer-centers Yes. Community oncology networks are a recognised segment with high patient volume and less integration engineering support than academic medical centers. The plug-and-play deployment uses SMART on FHIR for EHRs that support it, and clinical note paste or upload for those that do not. Care Journal patient self-reporting is a particularly strong fit for community settings where between-visit symptom monitoring drives both billing (RTM CPT codes) and earlier toxicity detection. ## Why do NCI-designated cancer centers buy Burna AI? Category: for-cancer-centers Three reasons. First, coordinator capacity: at a typical NCI-designated cancer center running 100+ concurrent trials, manual CTCAE grading at 15 to 20 minutes per adverse event consumes hundreds of coordinator hours every month. Burna AI compresses grading time by at least 70%, which translates into capacity for additional trials without additional headcount. Second, audit compliance: cited grades with WHO-UMC and Kramer attribution and complete audit trails reduce findings on grading consistency in sponsor monitor visits. Third, coordinator retention: shifting time from documentation to patient care is the single highest-leverage move on coordinator wellbeing and turnover. ## Can Burna AI run across multiple sites in the same trial? Category: for-cros-and-phase-1-sites Yes, by design. Multi-site trial cross-site consistency is a core architectural commitment. Every grading decision propagates in real time across all stakeholders; submission-gated blinding preserves grading integrity; cross-site visibility means a sponsor or CRO operations lead sees grading inconsistencies as they happen, not at the next monthly monitoring visit. Organisation-scoped access keeps each site's work isolated where it should be and visible where it needs to be. ## What is the primary pilot KPI for a CRO or Phase 1 site? Category: for-cros-and-phase-1-sites Query reduction. In Phase 1 oncology, every lab and every symptom can become a monitor query. When the grade lands in the EDC with the source sentence, the CTCAE criterion, the rationale, and the attribution support already attached, avoidable queries drop. Secondary metrics: EDC completeness, sponsor review turnaround, time to database lock. ## How does Burna AI serve CROs and Phase 1 sites? Category: for-cros-and-phase-1-sites CROs and Phase 1 sites manage high-frequency adverse event capture across multiple sites and multiple sponsors. Burna AI deploys across every site in a trial; grading inconsistencies surface in real time across sites, not at the next monitoring visit; every grade ships with the citation and attribution evidence already attached; and Medidata Rave or Veeva Vault EDC write-back closes the loop into the systems monitors and sponsors review. The primary success metric for this segment is query reduction, not initial grading speed. ## How does Burna AI address the 11% oncology death rate attributable to poor adverse event management? Category: for-pharma-sponsors Roughly 11% of deaths in oncology are attributable to poor adverse event management (industry estimate). Burna AI's leverage on that number is concentrated at three points. First, faster and more accurate attribution at the point of clinical encounter enables earlier dose modification and intervention. Second, continuous between-visit symptom monitoring through Care Journal captures toxicities that would otherwise present at the next clinic visit (Denis et al. JAMA 2019 showed a 22.5 vs 14.9 months OS benefit with structured patient self-reporting). Third, real-time cross-site visibility in multi-site trials means a safety signal in one site is visible to all stakeholders the moment it lands, not at the next monthly monitoring visit. ## How does Burna AI serve pharma sponsors? Category: for-pharma-sponsors Two product lines for pharma. Protocol Safe is a per-trial sponsor-tenant architecture for protocol-aware adverse event attribution where the protocol stays inside the sponsor's own cloud infrastructure; only the attribution signal crosses the boundary. Postmarket Pharmacovigilance Processing is an enterprise platform that uses the same grading engine to process adverse events on approved drugs, with MedDRA coding, E2B(R3) formatting, MedWatch and CIOMS generation, and DSUR and PSUR support. The throughline: the same engine, the same citation discipline, applied across the full drug lifecycle from Phase 1 trial to decades of postmarket surveillance. ## How does Burna AI support postmarket pharmacovigilance? Category: for-pharma-sponsors The same twelve-agent grading engine processes adverse events on approved drugs. MedDRA coding feeds the terminology layer. E2B(R3) formatting feeds electronic submission to regulators (FDA, EMA, PMDA). MedWatch forms feed US submissions; CIOMS reports feed international submissions. Periodic safety reporting (DSURs for development phase, PSURs for postmarket) is generated from the structured event store. Real-time case processing covers the 7-day and 15-day expedited reporting timelines. ## What is the postmarket pharmacovigilance volume scale Burna AI is built for? Category: for-pharma-sponsors FDA receives 2 million+ adverse drug reaction reports annually across all approved drugs. A large pharma company can process over a million ICSRs per year in-house. A single high-volume oncology drug can generate hundreds of thousands of postmarket ICSRs per year for the duration of its market life. Burna AI's postmarket module is architected for commercial-scale volume: real-time case processing, MedDRA coding, E2B(R3) submission output, and structured data feeding directly into safety signal detection. ## How does Protocol Safe preserve sponsor protocol IP? Category: for-pharma-sponsors By architecture, not by promise. The protocol PDF is ingested into a retrieval store that sits inside the sponsor's own cloud tenancy. The retrieval agent runs in that tenancy. The query Burna AI sends in is the de-identified adverse event narrative; the response is a structured per-drug attribution probability score. Content boundary middleware fail-closes against any response that contains protocol text. Burna AI cannot read the protocol, cannot inadvertently train on it, and cannot exfiltrate it. The sponsor IP stays inside the sponsor cloud at all times. ## What is Protocol Safe? Category: for-pharma-sponsors Protocol Safe is the pharma sponsor product line for protocol-aware adverse event attribution. An isolated AI environment is deployed inside the pharma sponsor's own cloud infrastructure. A retrieval agent ingests the protocol behind the sponsor's boundary and is queried for the trial drug's expected adverse event profile. Content boundary middleware fail-closes against any response that is not a structured probability score. Only the attribution signal crosses the boundary. The protocol never leaves the sponsor's vault. Self-hosted MedGemma handles attribution reasoning; BAAI/bge-large-en-v1.5 handles embeddings. ## How do you pronounce Nnenna? Category: founder-and-team Neh-NAH. Nigerian heritage. ## Who founded Burna AI? Category: founder-and-team Nnenna John, Founder and CEO. 19+ years of platform engineering experience, including 6 years at Airbnb building infrastructure for 150M+ users across 220+ countries. Prior roles at JP Morgan Chase, Roku, Expedia, and Stella & Dot. Atlanta-based. Nnenna personally architected and built the entire grading workflow, every agent step from raw phrase extraction through grading with citation. Paschal Ezeugwu is Lead Engineer and Co-Founder, with deep EHR expertise and ownership of clinical data ingestion. ## Who is on the Burna AI advisory board? Category: founder-and-team The active advisory board spans clinical oncology, pharma and clinical development, federal health IT, regulatory, and clinical informatics, with 20+ advisors. The list below names people who appear in operational contexts such as demos, regulatory discussions, and conference presentation. Named advisors (sample): - Dr. Aman Opneja, Chief Trials Officer and Clinical Advisor, multi-drug attribution and Phase 1 oncology trial design. - Dr. Mary Morison Saltz, CMIO, Stony Brook Cancer Center, data quality at scale and N3C-scale clinical informatics. - Dr. Stefan Gluck, oncology conference presentation across Europe and the United States, open to fractional CMO path. - Dr. Usman Shah, Medical Director, GI Oncology and Phase 1 Trials, Atlantic Health System, attribution at the point of clinical encounter. - Dr. Andrea Pirzkall, Oncology Clinical Development Consultant, former Genentech, BeiGene, Replimune, FDA regulatory strategy and immuno-oncology toxicity. - Dr. Azita Hamedani, President, UNC Health Faculty Physicians, system design and architectural framing. - Howard Fingert, FDA credentialing and regulatory pathway authority. - Joel Saltz, clinical informatics and data infrastructure authority. - Michel Azoulay, Strategic Advisor to the CEO, Protocol Safe pharma BD discussions. ## Why did Nnenna John start Burna AI? Category: founder-and-team Years of conversations with oncologists, research coordinators, and pharmacovigilance leaders surfaced a single architectural gap: oncology adverse event grading and attribution sit at the chokepoint of drug development, the inter-rater variability is well-published, the manual grading burden is concentrated, and no existing platform helps the investigator make the attribution decision at the point of clinical encounter. The Airbnb platform engineering experience is the foundation: every cancer trial deserves architecture that can defend its safety data, and citation-bound design is the only honest way to deploy LLMs into regulated clinical workflows. ## How does Burna AI compare to Flatiron Health as a business model? Category: market-and-investors Flatiron captured oncology real-world evidence from EHRs and sold the resulting dataset and analytics to pharma. Flatiron was pre-AI and single-sided: it served the data buyer (pharma), not the data producer (the oncologist), and was acquired by Roche for $1.9 billion in 2018. Burna AI is AI-native and two-sided. Provider SaaS at cancer centers and CROs (the data producer); Protocol Safe and Postmarket Pharmacovigilance at pharma (the data buyer); the data quality layer is the company; the product is the proof you can build the layer. Same Flatiron structural insight, applied to higher-stakes data (safety, not just outcomes) with native AI architecture and a substantially larger TAM across the drug lifecycle. ## Is Burna AI raising a round? Category: market-and-investors We are currently fundraising and bringing on a small group of investors before this stage closes in May. The Wefunder SAFE round page is at https://wefunder.com/burnaai.inc. The investor pitch deck is at https://burnaai.docsend.com/v/y3r53/pitch-deck. Happy to share details if it is interesting to you. ## What is the total addressable market for Burna AI? Category: market-and-investors $15 to $20 billion+ across the drug lifecycle, distributed across five layers. Clinical trial AE grading: $4 billion+ (manual grading at a significant share of clinical trial operating costs across thousands of global oncology trials). Patient self-reporting and RTM: $1 to $2 billion. Postmarket pharmacovigilance: $8 to $10 billion+ (global PV market; FDA receives 2 million+ adverse drug reaction reports annually, processing mandated by regulation for every approved drug, indefinitely). Safety signal analytics: $1 to $2 billion. Protocol Safe (pharma sponsor architecture): $500 million to $1.5 billion today, growing to $2 to $5 billion by 2030. ## Does Burna AI have a mobile app? Category: mobile-apps Yes. Two mobile apps. The provider app is a React Native application for clinicians and coordinators: CTCAE term search, manual grading, AI-assisted grading, encounter recording with audio, patient assignment, symptom management queue, dark mode parity, native iOS and Android. The patient app is a token-based, passwordless app for symptom self-reporting through Care Journal, available in seven languages. ## What can a clinician do in the Burna AI mobile provider app? Category: mobile-apps Search CTCAE terms across 1,000+ entries grouped by System Organ Class; view grade definitions; manually grade adverse events with full note-taking support; record clinical encounters with audio capture; run AI-assisted grading and see real-time workflow progress; review and approve AI suggestions; assign patients with EHR search and import; review the symptom management queue for patient-reported events; sign approvals with 21 CFR Part 11 e-signature. Full dark mode and light mode support; mobile-responsive across phone and tablet. ## How does Burna AI compute attribution? Category: multi-drug-attribution Attribution uses two named causality algorithms with visible math: the WHO-UMC (World Health Organization-Uppsala Monitoring Centre) classification and the Kramer algorithm. The engine produces per-drug probability scores (certain, probable, possible, unlikely, conditional, unassessable for WHO-UMC; and a multi-dimensional score for Kramer). The math is shown, not hidden. This is the opposite of a black box. The clinician reviews and decides; the architecture preserves the reasoning. ## How many oncology regimens does Burna AI cover? Category: multi-drug-attribution 42 pre-built oncology regimen profiles, covering common combination therapies including FOLFOX, FOLFIRI, carboplatin and paclitaxel, R-CHOP, ABVD, FOLFIRINOX, carboplatin and pemetrexed, checkpoint inhibitor combinations (pembrolizumab plus chemotherapy, nivolumab plus ipilimumab), antibody-drug conjugates, and CAR-T plus bridging regimens. Multi-drug attribution works out of the box without custom configuration. ## How does Burna AI handle protocols where the trial drug profile is sensitive sponsor IP? Category: multi-drug-attribution Through Protocol Safe, a sponsor-tenant architecture where the trial protocol stays inside the pharma sponsor's own cloud infrastructure. A retrieval agent ingests the protocol behind the sponsor's boundary and is queried for the trial drug's expected adverse event profile. Content boundary middleware fail-closes against any response that is not a structured probability score. Only the attribution signal crosses the boundary. The protocol never leaves the sponsor's vault. ## Can Burna AI grade single-agent trials too? Category: multi-drug-attribution Yes. Single-agent trials are a simpler case of the same engine. Attribution still runs through WHO-UMC and Kramer, but with one drug the probability scores collapse to the single agent. The grading, citation, and audit trail behaviour is identical to combination regimens. ## What is multi-drug attribution? Category: multi-drug-attribution Multi-drug attribution is the process of determining which drug or drugs in a combination regimen most likely caused a given adverse event. In oncology, combination therapy is the standard of care: FOLFOX, FOLFIRI, carboplatin and paclitaxel, checkpoint inhibitor combinations, antibody-drug conjugates, CAR-T plus bridging chemotherapy. When fatigue, neutropenia, or hepatic injury occurs, attributing it correctly is the foundation of dose modification, SAE narrative writing, and downstream pharmacovigilance signal detection. ## What are WHO-UMC and Kramer causality algorithms? Category: multi-drug-attribution WHO-UMC is the World Health Organization-Uppsala Monitoring Centre causality assessment system, an internationally accepted framework that classifies drug-event causality based on temporal relationship, response to dechallenge or rechallenge, biological plausibility, and alternative explanations. The Kramer algorithm is a structured scoring framework that scores six axes (prior experience, alternative causes, timing, drug levels, response to dechallenge, response to rechallenge) to produce an overall causality category. Both are peer-reviewed, regulator-recognised, and reproducible. Burna AI applies both in parallel and surfaces the per-drug probability scores for clinician review. ## How does Burna AI handle disagreements between raters? Category: multi-rater-quality Through an automated adjudicator workflow with collision detection. When two independent raters disagree on a CTCAE grade, the system automatically detects the conflict and assigns an adjudicator from a pre-designated pool. The adjudicator sees both grades side by side and resolves with one click. Batch adjudication lets the adjudicator resolve multiple disagreements in a single submission (40% faster than one-at-a-time). Every resolution carries e-signature support for 21 CFR Part 11 compliance. Digest notifications keep the team informed without inbox overload. ## What is the baseline inter-rater agreement in oncology adverse event grading? Category: multi-rater-quality Inter-rater agreement on adverse event attribution sits at a kappa of 0.59 to 0.68 (Hong et al., 2020). Attribution change rate between investigators and central review runs 31 to 36% (Hillman et al., JCO 2010). The FDA-NCI 2019 attribution workshop characterised the current state as sub-optimal, unreliable, and inefficient. These are the gap statements that motivate Burna AI's citation-bound architecture. ## How does Burna AI measure inter-rater reliability? Category: multi-rater-quality A six-tab analytics dashboard shows Cohen's kappa statistics, classification metrics, confidence calibration, weekly trends, agreement matrices, and override rates. Data streams in real time from grading events to the analytics layer, so site leadership has live answers to questions like 'how many cases were graded this week,' 'what is our inter-rater agreement rate,' and 'where is the AI suggestion being overridden most often.' Audit-ready output supports FDA expectations for documenting inter-rater reliability in AI/ML tool deployments. ## What is blind grading? Category: multi-rater-quality Blind grading is a workflow where two or more independent raters grade the same adverse event without seeing each other's assessments. It is the cornerstone of clinical trial quality assurance, inter-rater reliability measurement, and grader certification. Burna AI's blind grading workflow lets coordinators seed standardised test cases, assign graders in bulk, enforce submission-gated visibility (admins and adjudicators can only see a rater's work after that rater has formally submitted), and surface real-time status without exposing intermediate work. ## Is Burna AI a CancerX member? Category: partnerships-and-validation Yes. Burna AI is a member of CancerX, the public-private partnership convened by HHS and ONC under the Biden administration's Cancer Moonshot Initiative. CancerX gathers cancer centers, technology companies, federal agencies, and patient advocacy groups around the common goal of accelerating oncology innovation. ## Is Burna AI validated? Category: partnerships-and-validation Burna AI does not yet meet the regulatory threshold for the term 'validated,' which requires a completed clinical validation study with documented protocol, IRB review, statistical pre-registration, inter-rater comparison against expert clinicians, and peer-reviewed publication. The platform shows strong agreement with expert clinicians in ongoing internal testing. A 1,200-chart US accuracy validation study is in progress through Mayo Clinic Platform_Accelerate (June 2026 cohort). Peer-reviewed publication targets include JCO, JAMIA, and JCO Clinical Cancer Informatics. ## Does Burna AI have peer-reviewed publications? Category: partnerships-and-validation Peer-reviewed publication is the target for 2026 and 2027, alongside conference scientific posters. Bio-IT World 2026 is the scientific poster venue for the US accuracy study. ASCO and AACI CRI are the next academic venues. JCO, JAMIA, and JCO Clinical Cancer Informatics are the journal targets. ## Is there published evidence for Burna AI's approach? Category: partnerships-and-validation Yes, on the supporting clinical literature. Inter-rater agreement on adverse event attribution today: kappa 0.59 to 0.68 (Hong et al., 2020). Attribution change between investigator and central review: 31 to 36% (Hillman et al., JCO 2010). The FDA-NCI 2019 attribution workshop characterised the current state as sub-optimal, unreliable, and inefficient. Survival benefit with patient symptom self-reporting: 22.5 vs 14.9 months median OS (Denis et al., JAMA 2019). ER visit reduction with patient self-reporting: 34% vs 41% (Basch et al., J Clin Oncol 2016). Burna AI's own internal testing performance and the in-progress 1,200-chart accuracy study will be published through peer-reviewed venues. ## Who are Burna AI's design partners? Category: partnerships-and-validation Burna AI is in active design partnership conversations and pilots across NCI-designated cancer centers, academic medical centers, CRO networks, and pharma sponsors. Named, public references include Atlantic Health System (Technology Subcommittee demo April 2026, pilot under discussion), Memorial Sloan Kettering (direct leadership engagement on the scale of AE workflow at major cancer centers), Moffitt Cancer Center (9 of 10 rating from clinical research leadership), UPMC Enterprises (NDA signed, scoping in progress), Lexington Medical Center (active design partnership conversation). Burna AI is also a member of CancerX, the public-private partnership convened by HHS and ONC under the Cancer Moonshot Initiative, and is enrolled in Mayo Clinic Platform_Accelerate (June 2026 cohort, access to 1,200+ patient charts across three campuses). ## How long does a pilot take? Category: pricing-and-pilots Typical pilot duration is 2 to 3 months. Integration setup (SMART on FHIR or EDC) takes days to a week. The remaining time is structured grading, paired-grader timing studies, query reduction measurement (for CRO and Phase 1), and attribution consistency review (for pharma). Pilot completion produces a measurable outcome report that supports the enterprise procurement decision. ## How much does Burna AI cost? Category: pricing-and-pilots Pricing is under design partnership terms today and depends on segment. Cancer centers and academic medical centers engage through pilot pricing with enterprise expansion after a defined 2 to 3 month validation period. CROs and Phase 1 sites engage per-trial or under platform license. Pharma sponsors engage through Protocol Safe per-trial research agreements and Postmarket Pharmacovigilance Processing platform plus per-case usage agreements. Book a 15-minute conversation to discuss pricing relevant to your organisation: https://calendly.com/nnennaj/chat-with-nnenna-john ## How does a Burna AI pilot get scoped? Category: pricing-and-pilots Step one is a 15-minute conversation to understand your trial mix, EHR/EDC environment, and which buyer segment fits. Step two is a 60-minute discovery to define the pilot scope: one department or one trial workflow, named success metrics, named clinical and IT stakeholders, named timeline. Step three is integration setup (Burna team handles it). Step four is the structured pilot period. Step five is a pilot outcome review that converts into enterprise pricing if the agreed metric moves. ## What does Burna AI charge for a clinical trial deployment? Category: pricing-and-pilots Pricing depends on segment and is set under design partnership terms today. CRO and Phase 1 site engagements run per-trial or under enterprise platform license. Cancer center and academic medical center engagements run under pilot pricing with enterprise expansion after pilot validation. Pharma Protocol Safe runs per-trial research agreement. Postmarket pharmacovigilance runs platform fee plus per-case usage. Specific numbers are shared in discovery conversations. Book a 15-minute conversation: https://calendly.com/nnennaj/chat-with-nnenna-john ## What is the Design Partnership Program? Category: pricing-and-pilots A low-commitment pilot, typically 2 to 3 months, with one department, Phase 1 site, or CRO workflow. The Burna AI team handles SMART on FHIR or EDC integration. Success criteria are defined per partnership: CTCAE grading time reduction for cancer centers; query reduction and EDC completeness for CRO and Phase 1 sites; attribution consistency for pharma. If the agreed metric does not move, there is no obligation to continue. Design partners shape the platform and receive priority access to new capabilities. ## What is the published evidence for patient symptom self-reporting in oncology? Category: pro-ctcae-patient-reporting Patient self-reporting of symptoms during oncology treatment is associated with a 22.5 vs 14.9 months median overall survival benefit (Denis et al., JAMA 2019), a 31% vs 41% reduction in emergency room visits (Basch et al., J Clin Oncol 2016), and downstream reductions in hospitalisation and chemotherapy interruptions. The survival benefit is larger than most oncology drugs produce. Care Journal is built to operationalise this evidence inside structured CTCAE grading. ## How does Burna AI handle a Grade 3 or higher patient-reported event? Category: pro-ctcae-patient-reporting When a patient self-reports a Grade 3 or higher adverse event via PRO-CTCAE, the assigned provider receives an automatic notification with a time-based action deadline. The case lands in the Symptom Management queue in the provider app with the full grade proposal, source citation, and rationale. The clinician reviews, edits if needed, signs with 21 CFR Part 11 e-signature, and the event flows into the encounter record with full audit trail. ## How does Burna AI support patient-reported symptoms? Category: pro-ctcae-patient-reporting Through Care Journal, a token-based patient self-reporting mobile module with automatic CTCAE translation. Patients receive a passwordless enrollment token tied to a specific protocol, site, and cycle, then log symptoms from their phone with voice or text input. The PRO-CTCAE submission is routed through the same CTCAE Master Workflow that grades clinician-entered events: a clinical narrative is synthesised from the structured PRO responses, the engine produces a complete CTCAE grade proposal with citation, and the assigned provider receives a notification for Grade 3 or higher events with a time-based action deadline. The clinician reviews and signs with 21 CFR Part 11-compliant e-signature. ## What languages does the patient self-reporting module support? Category: pro-ctcae-patient-reporting Seven languages: English, Spanish, French, German, Italian, Chinese, and Japanese. Translation files are consolidated per language for consistent terminology and faster locale onboarding. ## What is PRO-CTCAE? Category: pro-ctcae-patient-reporting PRO-CTCAE stands for Patient-Reported Outcomes version of the Common Terminology Criteria for Adverse Events, developed by the National Cancer Institute. Where clinician-graded CTCAE captures the clinician's interpretation of an adverse event, PRO-CTCAE captures the patient's direct experience of frequency, severity, and interference with daily activities. PRO-CTCAE is the foundation of patient-reported safety reporting in modern oncology trials. ## What is RTM billing? Category: pro-ctcae-patient-reporting RTM stands for Remote Therapeutic Monitoring. CMS reimbursable CPT codes 98975 to 98981 cover the device setup, monthly monitoring, and clinician interaction associated with continuous patient-reported symptom monitoring. Care Journal is built to be RTM-billable, so continuous CTCAE monitoring between clinic visits creates a revenue line in addition to the cost-savings story. ## How does Burna AI describe its mission? Category: product-overview To build the safety data quality layer for clinical research, from early phase trials through decades of postmarket surveillance. The long-term vision is the Flatiron model applied to the highest-stakes data in drug development. Flatiron captured real-world evidence during active treatment and was acquired by Roche for $1.9 billion. Flatiron was pre-AI and single-sided. Burna AI is AI-native and two-sided: structured, citation-bound, attribution-scored adverse event data from Phase 1 trial through 20 years postmarket. ## What is the Burna AI philosophy? Category: product-overview Implementation over hype. The discipline is to do one thing and do it well: safety data quality with complete transparency and human oversight. Burna AI automates one specific workflow (adverse event grading and attribution) using named algorithms, citation-bound architecture, and mandatory human review. Features that break the architectural integrity have been deliberately removed: dose management was killed because of FDA SaMD risk; RECIST was killed because it breaks the same-engine logic. ## What is the Burna AI tagline? Category: product-overview The safety and data quality platform for clinical research, from early phase trials through postmarket surveillance. ## What category of product is Burna AI? Category: product-overview Drug safety technology, specifically AI for pharmacovigilance and oncology adverse event management. Burna AI is a clinical workflow and safety data quality tool, not a SaMD (software as a medical device) and not a diagnostic device. It does not prescribe dose modifications, does not assess tumor response (RECIST), and does not replace clinician judgment. Comparable category framing: a real-time, citation-bound substrate for the oncology safety data layer, applied across the full drug lifecycle from Phase 1 through decades of postmarket surveillance. ## What does Burna AI do in one sentence? Category: product-overview Burna AI turns unstructured clinical documentation into governed, audit-ready safety intelligence by extracting adverse events, suggesting CTCAE grades, scoring per-drug attribution with WHO-UMC and Kramer algorithms, and attaching mandatory citations to source text and CTCAE criteria for clinician review and approval. ## What is Burna AI? Category: product-overview Burna AI is the safety and data quality platform for clinical research, from early phase trials through postmarket surveillance. The wedge product is CTCAE AI, an architecture for adverse event grading and multi-drug attribution that is citation-bound by design, with outputs structurally constrained to valid CTCAE grades and mandatory citations to source text and CTCAE criteria. The grading engine is twelve specialized agents operating in a cascading constraint pipeline, where each agent's output bounds the valid output space of every downstream agent. Two patents filed. The same engine processes adverse events on drugs in clinical trials and on drugs already on the market, supporting pharmacovigilance across the entire drug lifecycle. Built for regulated environments from day one: HIPAA compliant, SOC 2 certified, 21 CFR Part 11 aligned. Human-in-the-loop, always. AI suggests, clinicians decide. At a glance: - The safety and data quality platform for clinical research - twelve specialized agents in a cascading constraint pipeline - Two patents filed on the architecture - WHO-UMC and Kramer causality algorithms for multi-drug attribution - 42 pre-built oncology regimen profiles, 26 organ system classes, 850 CTCAE criteria - HIPAA compliant, SOC 2 certified, 21 CFR Part 11 aligned - Human-in-the-loop, always. AI suggests, clinicians decide ## What problem does Burna AI solve? Category: product-overview Manual CTCAE grading takes 15 to 20 minutes per adverse event. Inter-rater agreement on adverse event attribution sits at a kappa of 0.59 to 0.68 (Hong et al., 2020). Attribution change between investigator and central review runs 31 to 36% (Hillman et al., JCO 2010). The FDA-NCI 2019 attribution workshop characterised the current state as sub-optimal, unreliable, and inefficient. Burna AI closes the gap with citation-bound architecture: every grade carries a source sentence, a CTCAE criterion, and per-drug attribution scores from WHO-UMC and Kramer algorithms. Coordinator time drops by at least 70%. Cross-site grading consistency surfaces in real time, not at the next monitoring visit. The clinician reviews and approves; the architecture preserves the audit trail. ## Who is Burna AI for? Category: product-overview Burna AI serves three primary buyer segments in oncology drug safety. First, Clinical Research Directors at NCI-designated cancer centers and academic medical centers managing 50 to 100+ active oncology trials. Second, CRO and Phase 1 site operators managing multi-site oncology trials where query reduction, EDC write-back, and cross-site visibility are immediate pain points. Third, VP Clinical Development, Head of Patient Safety, and Director of Pharmacovigilance at pharma and biotech sponsors that need protocol-aware attribution during trials and postmarket adverse event processing on approved drugs. ## Can Burna AI ingest a clinical trial protocol? Category: protocol-and-knowledge-base Yes. Upload a clinical trial protocol PDF and the platform extracts structured trial metadata (eligibility criteria, adverse event specifications, dose regimens), embeds the document for semantic search, and threads the protocol's own evidence into every CTCAE grading run on that trial. Coordinators review the extracted protocol structure in a dedicated UI before activating it. Once active, the protocol is the source of truth and the AI grading pipeline pulls criteria directly from it, with citations back to the original document. ## What are Clinical Preferences? Category: protocol-and-knowledge-base Institution-specific rules that guide AI grading across all workflow categories: CTCAE grading, drug attribution, drug interaction analysis, and encounter title generation. Configure once in Settings, and every AI suggestion follows your clinical protocols. Different institutions have different grading conventions and clinical priorities; Clinical Preferences ensure AI suggestions align with how your team practices, reducing review time and improving consistency across your organisation. ## What is the Knowledge Base feature? Category: protocol-and-knowledge-base Upload your organisation's protocols, drug labels, and clinical guidelines, and the grading engine automatically references them during CTCAE assessments. The system parses, indexes, and retrieves relevant context from your documents in real time, with retrieval-augmented generation grounded in your institutional knowledge. If no documents are uploaded, grading continues normally with zero disruption. ## How does Burna AI support 21 CFR Part 11 compliance? Category: regulatory-reporting 21 CFR Part 11 is the FDA regulation governing electronic records and electronic signatures. Burna AI is 21 CFR Part 11 aligned across four dimensions. Cryptographic signatures: all AI-generated grades and clinician approvals are signed with SHA-256, tamper-proof, and include complete provenance tracking. Audit trail: every action is logged with user, timestamp, IP, and resource, retained for 6 years. Access controls: organisation-scoped role-based access with adjudicator, grader, admin, and reviewer roles. System validation: the engine cannot produce a grade outside the defined CTCAE set or skip citation, enforced architecturally. ## Does Burna AI generate regulatory reports? Category: regulatory-reporting Yes. Three regulatory report types ship today: FDA MedWatch 3500A forms for serious adverse event reporting, IRB Unanticipated Problem reports, and Sponsor SAE reports. Each report type includes compliance tracking, validation rules, supplemental data collection, and a regulatory cascade flag so a single adverse event can be routed to multiple required submissions. Clinical-grade AE Report PDFs include severity-coloured cards, WHO-UMC and Kramer attribution categories, patient context, study site data, and summary statistics suitable for sponsor reporting and audit defense. ## Does Burna AI support E2B(R3) submissions? Category: regulatory-reporting Yes, through the postmarket pharmacovigilance module. E2B(R3) is the ICH electronic standard for individual case safety report (ICSR) submissions to regulators (FDA FAERS, EMA EudraVigilance, PMDA in Japan). The Burna AI postmarket module formats structured adverse event data into E2B(R3), generates MedWatch forms for US submissions and CIOMS reports for international submissions, and supports periodic safety reporting (DSURs for development phase, PSURs for postmarket). ## Does Burna AI support MedDRA coding? Category: regulatory-reporting Yes. The terminology resolution step in the grading pipeline matches against the MedDRA hierarchy (System Organ Class, High Level Term, High Level Group Term, Preferred Term, Lowest Level Term), with 79 HLT groupings spanning 458+ individual terms across 26 organ system classes. MedDRA coding feeds postmarket pharmacovigilance processing and E2B(R3) submission output. ## What regulatory reports does the postmarket pharmacovigilance module produce? Category: regulatory-reporting Individual case safety reports (ICSRs) in E2B(R3) electronic format, MedWatch 3500A forms for FDA submission, CIOMS I reports for international submission, DSURs (Development Safety Update Reports) for trials in active development, and PSURs (Periodic Safety Update Reports) for marketed drugs. Real-time case processing covers time-sensitive reporting timelines (7-day expedited for fatal or life-threatening unexpected; 15-day expedited for other serious unexpected). ## What is the Burna AI 90-day mission? Category: roadmap Build authority on AE attribution architecture in oncology drug safety through May to August 2026. The publishing topology covers a twice-weekly Substack publication (Oncology Safety Intelligence), daily LinkedIn cadence on personal and company pages, weekly customer-facing changelog on burna.ai/blog, daily SEO articles activating in June 2026, and conference presence at SCOPE X, Bio-IT World 2026, ASCO 2026, and AACI CRI 2026. ## What is on the Burna AI roadmap? Category: roadmap Three lanes. Same engine, new lane (therapeutic adjacency that preserves the CTCAE substrate): gene therapy long-term follow-up under the FDA 15-year mandate; cardio-oncology; rare disease oncology indications; expanded postmarket pharmacovigilance for approved oncology drugs. Same engine, new geography: EU and UK deployment, with EMA E2B(R3) submission support. New engine, new market (separate engine, deliberate sequencing): psychiatric AE grading, which has its own grading framework. Features deliberately not on the roadmap: dose modification AI (FDA SaMD risk, killed), RECIST and tumor response (breaks same-engine logic, killed), patient recruitment and matching (crowded, distinct buyer). ## What features were deliberately removed from the Burna AI roadmap? Category: roadmap Two features were removed by design. Dose management AI (chemo modification AI): removed because it carries FDA SaMD (software as a medical device) regulatory risk that would change the entire compliance posture of the platform. RECIST and tumor response assessment: removed because it breaks the same-engine logic, requires a separate framework, and serves a different clinical workflow. Naming what was deliberately removed is part of how Burna AI communicates discipline. ## Does Burna AI sign Business Associate Agreements? Category: security-and-compliance Yes. BAAs are signed with every customer who is a Covered Entity or Business Associate under HIPAA, and with every subprocessor in the Burna AI data pipeline. The standard BAA template is available on request and can be redlined as part of enterprise procurement. ## Is Burna AI a medical device? Category: security-and-compliance No. Burna AI is a clinical workflow and safety data quality tool, not a SaMD (software as a medical device). The platform suggests; clinicians decide. The architecture is human-in-the-loop, with mandatory clinician review and approval before any grade becomes a record. Adjacent features that would carry SaMD risk, such as dose modification recommendations, have been deliberately removed from the roadmap. ## Is Burna AI FDA-approved? Category: security-and-compliance Burna AI is not FDA-approved. Burna AI is eligible for the FDA Center for Drug Evaluation and Research (CDER) Emerging Drug Safety Technology Program (EDSTP); the application is planned for June to July 2026. EDSTP is the FDA pathway for collaborative review of emerging drug safety technology, the program's named scope. The platform is a clinical workflow tool, not a SaMD device, so the 510(k), De Novo, and PMA pathways are not applicable. ## Is Burna AI HIPAA compliant? Category: security-and-compliance Yes. Burna AI is HIPAA compliant across administrative, physical, and technical safeguards. PHI is encrypted at rest and in transit. Access is role-scoped and organisation-isolated. Audit logging covers every PHI access event with 6-year retention. PHI is explicitly excluded from system logs and AI training pipelines. Business Associate Agreements (BAAs) are signed with all subprocessors and are available to enterprise customers. ## Is Burna AI SOC 2 certified? Category: security-and-compliance Yes. Burna AI is SOC 2 certified, with the SOC 2 Type II audit cycle in progress. The Trust Services Criteria covered are Security, Availability, and Confidentiality. The report is available under NDA for enterprise procurement. ## How does Burna AI handle PHI in logs? Category: security-and-compliance PHI is never written to system logs. Automated PHI exposure prevention runs at the development hook layer, blocking commits that would introduce console output of patient data. Logs use non-PHI identifiers (record IDs, organisation IDs, timestamps) for operational traceability. The result is zero PHI in logs, enforced architecturally rather than by manual review. ## Where is Burna AI hosted? Category: security-and-compliance Burna AI runs on US-based cloud infrastructure with multi-region availability, US data residency by default, and customer-isolated tenancy at the data layer. Three deployment models are supported: standard cloud deployment for the platform SaaS, customer-isolated deployment for enterprise customers requiring dedicated infrastructure, and Protocol Safe deployment inside the sponsor's own cloud for pharma protocol-aware attribution where the protocol must never leave the sponsor's boundary. ## How often does Burna AI ship updates? Category: shipped-features Continuously, with a weekly customer-facing release rhythm. A typical week ships 7 to 15 PRs, 3 to 9 new features or improvements, several bug fixes, and infrastructure work that compounds. The weekly changelog covers everything visible to the customer; the weekly developer changelog covers what the engineering and design partner audiences want to see. ## What major features did Burna AI ship in 2025 and 2026? Category: shipped-features Burna AI shipped a complete CTCAE grading and attribution system across 2025 and 2026. The four entries immediately below split that scope into grading and workflow, integration and knowledge, patient reporting and analytics, and compliance, mobile, and public surfaces. Deep dives in this category: - What grading and workflow features did Burna AI ship in 2025 and 2026? - What integration and knowledge features did Burna AI ship in 2025 and 2026? - What patient, reporting, and analytics features did Burna AI ship in 2025 and 2026? - What compliance, mobile, and go-to-market surfaces did Burna AI ship in 2025 and 2026? ## What compliance, mobile, and go-to-market surfaces did Burna AI ship in 2025 and 2026? Category: shipped-features Six-year audit log retention; SHA-256 cryptographic e-signature for 21 CFR Part 11 alignment; React Native provider app with dark mode parity; React Native patient app with token-based passwordless enrollment; HIPAA compliance, SOC 2 certification, and 21 CFR Part 11 alignment; passwordless demo at demo.burna.ai; public marketing and trust surfaces at https://www.burna.ai. ## What grading and workflow features did Burna AI ship in 2025 and 2026? Category: shipped-features AI-assisted CTCAE grading with confidence scoring and rationale; multi-event grading (1 to 10 adverse events per case in one pass); multi-drug attribution with WHO-UMC and Kramer and 42 pre-built oncology regimen profiles; blind grading and grader roles; automated adjudicator workflow with collision detection; submission-gated blinding; comorbidity-aware grading; copy-forward (Documentation Quality Assessment) detection; self-improving terminology across 79 HLT groupings, 458+ terms, 26+ organ system classes; CTCAE v5.0 and v6.0 with 850 criteria. ## What integration and knowledge features did Burna AI ship in 2025 and 2026? Category: shipped-features SMART on FHIR for Epic, Oracle Health (Cerner), and Athenahealth, with 5,800+ Epic and 1,900+ Cerner endpoints managed at organisation level; patient and encounter import from the EHR; Medidata Rave and Veeva Vault EDC write-back; trial protocol ingestion with retrieval-augmented search; Knowledge Base RAG for institutional protocols and drug labels; Clinical Preferences for institution-specific grading rules. ## What patient, reporting, and analytics features did Burna AI ship in 2025 and 2026? Category: shipped-features PRO-CTCAE patient self-reporting through Care Journal in seven languages, with full CTCAE Master Workflow grading on submissions; provider notifications for Grade 3+ events; FDA MedWatch 3500A, IRB Unanticipated Problem, and Sponsor SAE report generation; clinical-grade AE Report PDFs with WHO-UMC and Kramer attribution; real-time analytics with Cohen's kappa, productivity, quality, and practice-wide views. ## What is the Burna AI changelog? Category: shipped-features A weekly customer-facing changelog published every Friday at https://www.burna.ai/blog, covering new features, improvements, bug fixes, and weekly impact metrics. The changelog leads with the user-visible benefit and the why-it-matters context, not the technical implementation. Subscribe by email through the Burna AI website or follow the LinkedIn company page for the weekly digest. ## What is the Burna AI alpha environment? Category: technology-and-platform alpha.burna.ai is an early-access environment for design partners and internal feature validation before production rollout. Authentication flows work seamlessly across alpha, staging, and production. Design partners can validate new features without affecting production workflows. ## What is the Burna AI demo environment? Category: technology-and-platform demo.burna.ai is the passwordless demo environment. Sign up with name, email, and phone, skip password setup, and land in a working demo organisation in seconds. Designed for sales conversations and partner walkthroughs. ## Can I see Burna AI in action? Category: technology-and-platform Yes. Three options. Book a 15-minute conversation for a guided walkthrough: https://calendly.com/nnennaj/chat-with-nnenna-john. Self-serve into the demo environment at demo.burna.ai with passwordless sign-up. Or request a live demo through the request-demo page on https://www.burna.ai. ## How does Burna AI handle workflow failures? Category: technology-and-platform Workflow resumption. If a network hiccup or temporary failure interrupts a grading, attribution, or interaction workflow mid-run, the engine resumes from the failed step rather than re-running the entire workflow. A 'Retry Workflow' button appears in the error state and picks up where the workflow left off, without losing completed steps. For long-running workflows that exceed a 5-minute polling timeout, the UI guides a graceful retry rather than spinning indefinitely. ## Is Burna AI built for real-time use? Category: technology-and-platform Yes, by architecture. Every grading decision, every attribution score, every clinical alert propagates instantly across all stakeholders. The platform is built on real-time architecture from day one, not real-time wrappers retrofitted onto a batch substrate. In a multi-site trial, grading inconsistencies surface in real time across sites, not at the next monthly monitoring visit. Sponsors get direct AE visibility rather than the typical monthly or quarterly safety report.