Alera: Home Health Operations Platform
HIPAA-compliant platform for home health agencies. AI-powered patient onboarding, clinician scheduling, and audio-to-SOAP note generation.
TypeScript React tRPC Supabase Anthropic Claude Deepgram
The Problem
Home health agencies run on spreadsheets, phone calls, and disconnected EMR systems. Patient onboarding takes hours of manual data entry. Clinicians spend more time on documentation than care. Scheduling is a whiteboard.
Who It Helps
- Home health agency operators managing 50+ clinicians
- Clinical directors drowning in scheduling complexity
- Compliance officers who need audit trails without manual tracking
What I Built
A full-stack operations platform that handles the operational core of running a home health agency.
- AI patient onboarding — Extracts structured patient data from EMR documents using Claude. What took hours now takes minutes.
- Clinician scheduling — Constraint-based scheduling that accounts for certifications, geography, availability, and patient needs.
- Audio to SOAP notes — Clinicians record visit audio. Deepgram transcribes. Claude generates structured SOAP notes. Clinicians review and approve—human always in the loop.
- Multi-tenant architecture — Row-level security in Supabase. Each agency sees only their data. Built for compliance from day one.
System Design Highlights
- Stack: TypeScript, React, tRPC, Supabase (Postgres + RLS), Vercel
- AI: Claude for document processing and note generation, Deepgram for speech-to-text
- Security: Multi-tenant row-level security, encrypted at rest, audit logging
- Architecture: Type-safe end-to-end with tRPC. No REST guessing games.
Workflow
- Agency uploads EMR documents for new patient
- Claude extracts structured data (demographics, conditions, medications, care plans)
- Clinician reviews and confirms extracted data
- System schedules visits based on care plan, clinician availability, and constraints
- Clinician records visit, audio transcribed, SOAP note generated, clinician approves
Risks & Mitigations
- AI extraction accuracy: Human review required for all AI-generated content. Nothing auto-publishes to patient records.
- HIPAA compliance: Multi-tenant RLS, encrypted storage, audit trails. Architecture designed for compliance—but compliance is ultimately the agency’s responsibility.
- Scheduling complexity: Constraint solver handles the combinatorics; edge cases escalate to human schedulers.
Current Status
In active development. Core platform operational. Contact me for a current demo.
Building something in healthcare? Let’s talk.