Documentation, billing, and EMR sync — automated for physicians
MedicalSuite is the platform behind ACM Automated Scribing Services. I founded the company and built the system end to end — API services, data model, auth, scheduling infrastructure, and the web app — for skilled-nursing physicians. It's in paid daily use.
- encounters documented to date
- 7,000+
- clinician hours saved (estimated)
- 800+
- modules specified in the architecture
- 180+
- from zero to production, solo
- 6 months
What I built
One pipeline for documentation, billing, and EMR sync
A monorepo unifies clinical documentation, billing, and EMR sync on a shared data pipeline, specified across 180+ modules — so a single encounter flows through every stage without being re-entered.
LLM output behind a deterministic state machine
Azure OpenAI (GPT-4o) pipelines draft notes and billing codes, but every step runs behind a deterministic state machine — so the model's output stays verifiable instead of free-form.
Clinician review in a few clicks
The clinician dashboard is built around a review-and-approve workflow with diagnosis tag pickers and per-provider verification controls, cutting review down to a few clicks per note.
A job system that recovers on its own
A fault-tolerant APScheduler job system handles multi-profile execution, automated retry and recovery triage, and watchdog monitoring — the platform runs daily without someone babysitting it.
How an encounter flows
Encounter
Enters the shared data pipeline — one monorepo, 180+ modules.
Documentation
LLM draft (Azure OpenAI) → deterministic state machine.
Clinician dashboard
Review → tag diagnoses → approve, with per-provider verification controls.
Billing
Billing-code generation → the same state-machine gate.
EMR sync
Scheduled jobs → retry / recovery triage → watchdog monitoring.
Tech Stack
Why there's no public demo
MedicalSuite handles clinical data, so the codebase and the running system are private. I'm glad to walk through the architecture, the state machine design, and the trade-offs in an interview.
Ask for a walkthrough