MedAssist

A full-stack AI medical assistant scoped exclusively to medical inquiries, integrating the Claude API with a retrieval-augmented generation pipeline to produce evidence-grounded responses backed by PubMed literature.

Live at medassist-as.vercel.app

MedAssist interface
System
Full-stack AI medical assistant
Domain-constrained to reduce hallucination risk
Evidence source
PubMed literature
RAG pipeline retrieves peer-reviewed research to ground each response
Model
Claude API + RAG
Citations surfaced alongside every generated answer

The problem

General-purpose LLMs hallucinate medical information at unsafe rates
Responses lack citations, making them unverifiable in clinical contexts
Accessing relevant biomedical literature is slow and requires expertise

Approach

Constrain generation to domain-relevant PubMed sources
Retrieve supporting literature before generating a response
Surface citations so users can verify claims directly

Technical build

Claude API for natural language generation and reasoning
Vector search over an indexed PubMed corpus
Full-stack web application deployed on Vercel

Why it matters

Demonstrates responsible AI design in a high-stakes domain
Makes peer-reviewed medical knowledge more accessible
Reduces hallucination risk through retrieval grounding

Technology stack

Claude API Retrieval-Augmented Generation PubMed Vector Search Full-Stack Web Vercel