Every telemedicine visit produces a transcript. Almost no one turns that text into a healthcare decision.
What it is
Dr. JON is the clinical intelligence platform we built as an in-house ANY tech product. A physician sees a patient over video, the audio is transcribed (AssemblyAI), and an extraction layer that combines Claude, GPT-4o and Gemini returns structured data: likely ICD codes, prescriptions, the patient journey, risk signals. That data feeds a 7-page executive dashboard that lets a healthcare network run epidemiological analysis at scale.
The market problem
Telemedicine became a commodity in 2020. What never became a commodity was extracting analytical value from what happens inside the consultations. Today, healthcare networks buy care volume. They don't buy intelligence about it.
The approach
- Webfront (Next.js 13 + Radix) · doctor-patient chat with native recording
- Orchestration API (Node + Express + TypeORM + PostgreSQL) · clinical ETL, anonymization, S3 integration
- AI layer · Claude for clinical entity extraction, GPT-4o for cross-checked confidence, Gemini for long cases, AssemblyAI for audio
- Dashboard (Next.js 16 + React 19 + D3 + Recharts) · epidemiological BI across 7 views
- Videos (Remotion 4) · automated generation of institutional and promotional material
What changed
- 4,049 consultations processed with 82% average extraction confidence
- 312 active physicians, 1,840 unique patients, 28 connected institutions
- 74% average documentation quality, a new metric for the industry
Consolidated stack
Next.js 13/16 · React 18/19 · TypeScript · TailwindCSS · Express · TypeORM · PostgreSQL · MongoDB · Claude · GPT-4o · Gemini · AssemblyAI · AWS S3 · Stripe · Twilio · Vercel · Remotion 4