For clinics
Every consultation, coded and claim-ready — before the patient leaves the room.
One API call turns a consultation recording into a structured SOAP note, evidence-linked ICD-10 code candidates and NHIS claim items with denial-prevention checks. In English, Yoruba, Hausa, Swahili and Pidgin — offline-first, and always approved by the clinician.
From consultation to claim-ready, in four steps
The same pipeline powers the GELYPTIC MD clinician app and the public API — so your integration behaves exactly like the product your clinicians already trust.
Clinicians record in the GELYPTIC MD app or your own product uploads audio to the API. Capture works offline, and audio is deleted once transcription finishes.
English, Yoruba, Hausa, Swahili and Pidgin, with code-switching between them. Amharic and Oromo are in beta. Regional dialects are supported, with ongoing improvement.
The engine drafts a clean, structured SOAP note — Subjective, Objective, Assessment, Plan — with no invented facts. The clinician reviews and approves before anything leaves the draft stage.
ICD-10 code candidates are matched from a curated code table — never invented by the model — and NHIS and NHIF claim items are proposed with readiness checks that catch common denial reasons.
What the integration looks like
Submit audio, get an acknowledgement immediately, and receive the finished result on your webhook — signed so your server can trust it.
# 1. Submit the consultation audio — the job is queued immediately
curl -X POST https://gelypticmd.com/api/public/v1/transcription \
-H "Authorization: Bearer $GELYPTIC_API_KEY" \
-F "audio=@consultation.webm"
# -> 202 Accepted
# { "job_id": "job_9f2c4b1e", "status": "queued", "request_id": "req_..." }
# 2. Your server receives the finished result on your webhook,
# signed with your webhook secret so you can verify it:
{
"event": "note.completed",
"note": {
"soap": {
"subjective": "...",
"objective": "...",
"assessment": "...",
"plan": "..."
},
"icd10_candidates": [
{ "code": "E11", "label": "Type 2 diabetes mellitus",
"evidence": ["type 2 diabetes", "raised fasting glucose"] }
],
"billing_items": [
{ "scheme": "NHIS", "label": "General outpatient consultation" },
{ "scheme": "NHIS", "label": "Fasting blood glucose" }
],
"claim_checks": [
{ "label": "Indication documented for lab order", "passed": true }
]
}
}- Scoped API keys
- Least-privilege, named and revocable. A key can narrow the application's scopes but never widen them.
- Async jobs
- Audio submission returns 202 with a job id — your system never waits on a transcription.
- Signed webhooks
- Every delivery is HMAC-signed with your webhook secret and carries a request id for tracing.
- Versioned & additive
- The API is versioned under /v1/ and only ever gains capabilities additively — integrations never break.
What your clinic gets from every consultation
One approved note becomes everything the clinic needs downstream — documentation, coding, billing and interoperable exports.
A clean, sectioned SOAP note generated from the consultation, ready for the clinician's review and sign-off — in the clinic's language mix.
Every candidate is matched against a curated ICD-10 table and carries the exact note phrases that produced it, so coders can verify at a glance. Codes are never generated freely by the model.
Proposed billable items tagged by scheme — NHIS Nigeria, NHIF Kenya, private HMO or cash — with notes on bundling and what each scheme normally covers.
Deterministic checks that flag the usual denial reasons — a missing indication, a test inside a bundled package, a referral without an authorisation code — before the claim is filed.
Referral letters, clinical summaries and discharge-ready text drawn from the same approved note, so downstream documents stay consistent.
Approved notes export as FHIR R4 Bundles and HL7 v2.5.1 ORU^R01 messages, so results flow into the EMR systems clinics already run.
Coding your biller can actually trust
Most AI scribes ask a language model to guess codes. GELYPTIC MD treats coding as a lookup problem — which is why our candidates are verifiable.
The curated ICD-10 table spans primary care, tropical and outbreak diseases, maternal and newborn care, mental health, chronic disease and social-drivers-of-health codes. It is a maintained starter set — new codes are added continuously, and your biller always sees the evidence behind every candidate.
- Candidate codes matched only from the curated table — never generated by the model
- Note phrases shown as evidence for every candidate
- Clinician or biller confirms before anything is billed
- Codes come from a curated table
- ICD-10 candidates are matched programmatically from a maintained code table. The language model never invents a code.
- Every code shows its evidence
- Each candidate carries the exact phrases in the note that produced it, so your biller can verify the match in seconds.
- Claims are proposals, not submissions
- Claim items and readiness checks are suggestions for the clinician or biller to confirm. GELYPTIC MD never submits a claim automatically.
- Clean claims, fewer denials
- Readiness checks flag missing indications, bundled items that should not be billed separately, and referrals without authorisation codes before filing.
Built for real-world clinics, not demo rooms
Low-connectivity rooms, multilingual consultations, staff who bill at the end of a long day. This is who we built for.
Built for consultations that switch between English and Yoruba, Hausa, Swahili or Pidgin mid-sentence — plus AAVE, Spanglish and regional Arabic and Spanish dialects.
Encrypted on-device recording with background sync. Consultations are captured with or without internet and queued for processing when connection returns.
Nothing is filed or exported without clinician approval. Coding and billing output are candidates to confirm — never auto-submitted on the clinic's behalf.
Large batches and busy queues never hang your system. Work runs as queued jobs and results arrive on signed webhooks — your app responds in milliseconds.
Clinic questions, answered
The questions clinic leads and practice managers ask us most.
Put your consultations to work.
Start free in the clinician app, or integrate the API into your own product. No credit card, no install, no audio ever stored.