Regulated conversations, signed.
Advisory boards, pharmacovigilance review, regulatory pre-submission meetings — Felarity produces tamper-evident records. Every transcript is diarized, every contradiction is attributed to a speaker, and every report is signed with Ed25519 and chained into an 8-node SHA-256 Merkle attestation so the provenance survives the meeting.
What pharma teams use Felarity for
- Advisory board meeting analysis. Diarized and signed transcripts for submission packaging. KOL discussion gets captured with speaker-level attribution so when Medical Affairs revisits a statement six months later, the record shows who said it, when, and in what context.
- Pharmacovigilance review. Contradictions across investigator reports — when one site's safety narrative says one thing in a teleconference and the written report says another, Felarity surfaces the gap with the source utterances and timestamps attached.
- FDA / EMA pre-submission meetings. A chain-of-custody record for submission preparation. The signed report, the audio segments referenced, and the attestation manifest are exportable as a single packet that downstream regulatory affairs can append to the submission file.
- Internal CDS / DSMB reviews. A structured record for sponsor-IRB packets. Clinical Development Subcommittee and Data Safety Monitoring Board sessions produce a credibility-weighted narrative with speaker arcs, confrontation moments, and the council's adjudication of any contested claims.
The Pharma council
Five specialists analyze every session. They are not generic chat models — they are domain-shaped analyst roles that produce structured findings, then debate them against each other before synthesis.
Tracks claim consistency across investigators, the alignment of clinical narratives to published evidence, and the framing of off-label discussion. Flags statements that would need substantiation in a label or a publication.
Surfaces causality assertions, seriousness reclassifications, and contradictions between verbal report and written CIOMS / E2B narratives. Notes when a meeting statement implies a reporting obligation that has not yet been satisfied.
FDA-shaped analyst. Tracks whether commitments made in a pre-submission discussion match the sponsor's filing strategy, and whether agency questions have been answered or deflected.
EMA / CHMP-shaped analyst. Same role on the European side: rapporteur interactions, scientific advice, and Type II variation discussions get analyzed against EU-specific procedural expectations.
Reads the numerical and methodological claims in the room. Catches the difference between "trend toward significance" and a defined alpha threshold, flags multiple-testing exposure that the protocol does not cover, and challenges any statement of effect size that the underlying analysis does not support.
HIPAA + GxP posture
We sign a HIPAA Business Associate Agreement on Professional and Enterprise tiers; the BAA is available on request before contract. Under the GDPR we are the data processor and standard contractual clauses are included in our Data Processing Addendum. We do not currently make GxP-validation claims — formal computer-system-validation work is on our roadmap for Q3 2027, and we will publish the validation package when it lands. Until then, we recommend Felarity for non-GxP-critical processes: advisory boards, scientific exchange, internal review meetings, regulatory preparation sessions, and pharmacovigilance triage. We are explicit about this because the alternative — overclaiming validated status — is what gets sponsors into trouble with their own QA function.
Sponsor-IRB workflows
The signed report PDF and the attestation chain JSON go straight into the submission packet. Each export carries the Ed25519 signature, the public verify URL, and the 8-node Merkle root so an IRB reviewer — or an auditor years later — can confirm the file has not been altered since the meeting closed. The verification endpoint is public; the content of the meeting is not. Reviewers see only what the sponsor sends them, but they can independently confirm authenticity without needing access to our infrastructure.
Workflow example
A mid-stage oncology sponsor runs a Type B pre-submission meeting with the FDA division. The session is captured in Felarity with diarization on; the Medical Affairs Lead and Regulatory Affairs (US) analysts mark a moment where the agency reviewer asks for an additional confirmatory analysis, and the sponsor's response is partial. The contradiction-detection pass flags it during the live session. After the meeting, the eleven-stage post-session pipeline runs:
- Audio chunks are concatenated and diarized end-to-end.
- Acoustic markers — speech rate, pause patterns, stress — are extracted per speaker.
- Confrontation moments are classified as resolved, deflected, or unresolved.
- Voice samples are extracted for each speaker so attribution can be checked.
- The DeBERTa-v3 NLI re-scoring pass confirms which claims actually contradict.
- The topology pass classifies the pattern of disagreement across the session.
- The council runs a deep re-analysis with full speaker context.
- The 8-node SHA-256 Merkle attestation chain is sealed and Ed25519-signed.
The output: a signed PDF report, an attestation manifest, and the audio bundle. The regulatory affairs team appends the report to the internal meeting minutes folder, and the sponsor's commitments tracker is updated with the unresolved-confrontation item. When the formal minutes arrive from the agency three weeks later, Felarity's record is the reference document — not because we replace the agency's minutes, but because our record was sealed at the moment the meeting ended.
What we are not
We are not a transcription service. We are not a meeting summarizer. We are forensic intelligence over conversations that matter — and in pharma, conversations matter because the regulatory record built on them is what the agency, the auditor, and (occasionally) opposing counsel will read.