MedPath
Team consisting of Amedeo Ercole (UMD; Google‑backed RL researcher and FDSE at RGB HK — Bayesian RL, GPUs) and Olivia Moody (NYU; server engineer intern).
YouTube Video
Project Description
MedPath: Healthcare Agent Demo utilizing Cloud Run GPUs, BigQuery, and Gemma
MedPath is an AI-powered web app that helps users find the best healthcare providers based on their insurance, care type, and location — without needing to navigate complex directories or search portals.
Users enter their basic information, and MedPath does the rest: querying real-time web data, filtering by coverage, and summarizing the top 3 care options using a language model.
Prerequisites
Node.js (v18+ recommended)
npm or yarn
Internet connection (to access Cloud Run and fetch model results)
Backend Cloud Run endpoint deployed and accessible (update if using your own)
Setup
git clone https://github.com/livmood/MedPath.git
Run npm install or yarn install
Run npm run dev or npm run build
Features
Uses a local LLM (Gemma via Ollama) to interpret and summarize search results.
Pulls real-time provider options from Google via the SerpAPI.
Filters for in-network providers.
Avoids suggesting portals — it finds real providers, ready to serve.
Connects a modern frontend (React + Vite) with a Python-based agent backend via an API.