FHIRBrush - AI Tinkerers - New York City Hackathon
AI Tinkerers - New York City
Hackathon Showcase

FHIRBrush

Resolve complex technical issues locally with our RAG-driven autonomous agent, ensuring total privacy and human-in-the-loop security.

2 members Watch Demo

FHIRBrush is the ultimate way to visualize clinical constellation in a dynamic, changing patient condition. It turns real patient health records into a living, interactive, reasoning-driven node graph, powered by Claude and FHIR R4, the same data standard used by every major US hospital system.
There is no chat box. Clinicians watch the canvas reorganize itself in real time as new lab results stream in. For instance, when creatinine rises above threshold, Claude reasons over the full patient context and responds by highlighting at-risk nodes, drawing causal edges between related findings, and updating a one-sentence clinical narrative. The interface is not a window into Claude’s output. The interface is Claude’s output.
A 20-patient priority bar sorts patients by severity (red, orange, green) at a glance. Click any dot to jump into that patient’s full canvas instantly.

AI Tinkerers Anthropic Anthropic Claude 3.5 Sonnet Betaworks FHIR R4 FastAPI LOINC/ICD-10 clinical codes Python asyncio React 19 React Flow 11 Redis Redis. TypeScript Vite WebSockets