Projects
Major work organized by domain. Each project shows working artifacts and live demos.
HeidiChat - Clinical AI System
Project ID: HCHAT-2025-001 | Status: 🟢 Active Development
What This Is
Agentic AI that prepares patients for medical appointments. It injests medical data, interviews patients about symptoms, retrieves relevant history, and generates briefing docs for clinicians.
How It Works
[Interactive architecture diagram]
Click each component to see details ↑
Why I Built This Way
- → Chose RAG over fine-tuning (better for clinical accuracy)
- → Vector DB for semantic retrieval (ChromaDB)
- → Agentic workflow for complex reasoning (LangGraph)
What I Learned
RAG architectures need careful prompt engineering. Vector similarity alone isn't enough - you need semantic chunking and metadata filtering. Agentic workflows add latency but dramatically improve output quality for complex medical reasoning.
Feather Platform - Social Health Tech
Recommendation engine connecting people to group fitness. 1,700 users, 120% lift in social interaction, $500k raised.
Built with: React, Node.js, PostgreSQL, ML recommender
View case study →COVID-19 Modeling - Epidemic Response
Agent-based SEIR modeling that informed Australia's elimination strategy. Published research, thousands of lives saved.
Built with: Python, R, Agent-based modeling
See methodology →