ai-ml
GlucoRAG
Guideline-Grounded RAG Assistant — Python, FastAPI, Vite, Docker, AWS
PythonFastAPIViteDockerAWSGroqFastEmbedRAG
Highlights
- Built a retrieval-augmented generation pipeline delivering citation-backed diabetes screening answers grounded in Diabetes Canada and NICE NG28 clinical guidelines, using Groq for chat and local FastEmbed embeddings with a numpy cosine-similarity index.
- Implemented a FastAPI backend (
/query,/health) and a Vite/TanStack frontend, containerized with Docker and deployed via Vercel (frontend) and AWS (API). - Designed safety guardrails that refuse medication advice and personal diagnosis, validated by a custom Day 4 risk and claim-support evaluation suite.