Inspiration Healthcare providers spend 50% of their time on documentation instead of patient care. We built babelfhir-vibe to eliminate this burden using AI agents that transform clinical notes into structured FHIR data automatically.
What it does Instant FHIR Conversion: Transform unstructured clinical notes into HIPAA-compliant FHIR resources in seconds Multi-Agent Orchestration: Airia coordinates PhenoML (clinical NLP), babelfhir-ts (validation), and OpenAI (reasoning) in parallel workflows Real-Time Monitoring: Datadog tracks every API call, workflow step, and error across all 6 AI platforms Production Analytics: ClickHouse delivers sub-20ms queries on workflow performance metrics
How we built it PhenoML for medical coding & clinical NLP (Lang2FHIR, Construe) Airia for multi-agent DAG orchestration with parallel execution babelfhir-ts for type-safe FHIR validation (open-source NPM package) Datadog for distributed tracing, RUM, error tracking, and custom metrics ClickHouse for real-time OLAP analytics on workflow performance OpenAI as clinical reasoning fallback Next.js 15 + Auth0 for production-ready web app
Challenges we ran into Coordinating 6 different AI platforms with different APIs and rate limits Ensuring HIPAA compliance while maintaining full observability Building distributed tracing across async multi-agent workflows Real-time analytics without sacrificing performance
Built With
- airia
- babelfhir-ts
- clickhouse
- datadog
- freepik
- openai
- vercel
Log in or sign up for Devpost to join the conversation.