Atlas

Built by Raphael San Andres

An AI agent for exploring and analyzing FHIR clinical data. Ask questions in natural language and watch the RAG pipeline retrieve, reason, and respond.

91K patients
7.7M vectors
20+ medical tools
~$120/mo on AWS
Python
LangGraph
pgvector
AWS ECS
Claude 3.5
Next.js
Go
GitHubTechnical Docs
All patient data is synthetic, generated by Synthea. No real patient information is used.
How It Works
1
Ask a Question
Natural language query about any patient's clinical records
2
Multi-Agent RAG
Hybrid vector search, cross-encoder reranking, and LLM reasoning with 20+ medical tools
3
Cited Response
Grounded answers with source chunks, SNOMED codes, and tool citations
AWS architecture diagram
System Architecture

Multi-agent LangGraph pipeline on AWS ECS, RDS pgvector, and Bedrock Claude. Includes data flow, database schema, and retrieval pipeline diagrams.

Read the technical docs →
Select a Patient
15 featured