Case study · 09
ARIA recovery monitoring
A voice-first post-op recovery companion that streams sensor data and emits FHIR bundles for the EHR.
- Role
- Healthcare AI hackathon build
- Status
- Hackathon, complete
- Read
- 1 min
- Stack
- FastAPINext.jsWebSocketsWhisperClaude APISQLite
FHIR R4
bundles with LOINC and ICD-10 coding
Problem
Post-op recovery data (steps, range of motion, pain) rarely reaches clinicians in a usable form.
Approach
- 01A voice pipeline from microphone to Whisper to an LLM to the database.
- 02A persona engine that adapts tone from a 5-question onboarding survey.
- 03A real-time sensor stream with an alert engine, over WebSockets.
- 04A FHIR R4 generator mapped to LOINC and ICD-10, with a provider dashboard and an Epic-style bundle viewer.
From a patient's voice to a coded bundle on the provider dashboard.
Results
- A working end-to-end demo, from a patient's voice to a coded bundle on the provider dashboard.
What I'd do next
- Validate the coding against a real EHR sandbox.