Proof for every clinical claim your AI makes.
The enterprise verification layer for medical AI. ProofMed grounds any model's output against live authoritative sources, returns per-claim verdicts with citations, and seals every result into a tamper-evident evidence ledger.
A verification layer for medical AI
Everything you need to make clinical AI output auditable, cited, and safe to act on.
Grounded verification
Extract clinical claims from any text and check each against live authoritative databases — supported, contradicted, or unverifiable, with sources.
Tamper-evident ledger
Every verification is hash-chained and HMAC-sealed into an append-only provenance ledger — exportable as a signed evidence package for audit.
Multi-agent orchestration
Route, compress, and cross-check with a consensus engine and clinical judge — frontier reasoning at a fraction of the token cost.
FHIR R4 ingestion
Point a resource or Bundle at ProofMed and get grounded, cited proof for EHR data — verification at the point of care.
BYOK & tenancy
Bring your own model keys. Per-tenant isolation with row-level security, budget guards, and usage metering built in.
Sovereign fine-tuning
Launch clinical fine-tuning and alignment runs on fractional GPUs — HIPAA de-identification and ontology alignment by design.
From raw output to signed proof
Send any clinical text
Your model's output, a third-party model's, or a FHIR resource — one API call.
Ground & score
Claims are checked against live sources and scored for clinical fidelity, with citations attached.
Seal & export
The result is sealed into the ledger and returned as a signed, auditable evidence package.
One endpoint. Model-agnostic.
POST /v1/verify
{ "text": "Metformin is first-line for type 2 diabetes." }
# → per-claim verdicts · citations · clinical fidelity score · signed receipt
See the proof for yourself.
Sign in to run a verification, inspect the evidence ledger, and explore the platform.