Inspiration

Hospitals lose $157 billion annually from missed contract deadlines—equipment auto-renews at bad rates, insurance underpays, and research grants get cancelled. In 2025, NIH terminated $3.8 billion in research grants, many because hospitals simply missed reporting deadlines. Harvard lost $110 million, Duke lost $96 million, and cancer trials stopped mid-study. I realized teaching hospitals needed an AI assistant that never forgets a deadline.

What it does

ContractIQ reads contracts instantly using AI and helps administrators track deadlines via web dashboard or WhatsApp. Upload any contract via photo or PDF → AI extracts key dates, amounts, and terms → Dashboard shows everything color-coded (🔴 urgent, 🟡 soon, 🟢 safe) → Ask questions in plain English like "What expires in Q2?" → You can use ContractIQ entirely on WhatsApp too—upload contracts by photo, query details on-the-go, and get proactive alerts, all without opening a laptop. It tracks both vendor contracts (MRI leases, suppliers) AND research grants (NIH funding, clinical trials) in one place.

How I built it

I combined Claude AI via Model Context Protocol (MCP) for understanding contracts, local AI models for sensitive hospital data, and vision AI for reading scanned documents. Built a React dashboard with real-time chat, connected Flask backend to Twilio for WhatsApp integration, and used Supabase for secure database storage. The MCP server I built routes between local AI (privacy) and Claude API (power) based on document sensitivity.

Challenges we ran into

Railway deployment was brutal: My app's port settings conflicted with Railway's dynamic ports, I had to manually configure 15+ environment variables, builds failed due to Node version mismatches, and CORS errors blocked frontend-backend communication.

Accomplishments that I'm proud of

Created WhatsApp integration that lets administrators upload contracts and get alerts without a computer—solving real accessibility problems. Designed an intuitive red/yellow/green urgency system that requires zero training. Built an AI-powered extraction system that reads any contract format instantly—PDFs, scans, or photos. Most proud that I'm solving a $157 billion problem with technology that's 10x cheaper ($5K/year vs. $100K+) than existing enterprise solutions.

What I learned

Deployment platforms like Railway require completely different configurations than local development—you can't just "push to production." Combining multiple AI models is architecturally complex but creates way better user experiences. WhatsApp integration dramatically increases accessibility for mobile professionals. Real impact comes from deeply understanding problems: hospitals don't need fancy features, they need to never miss another deadline. Sometimes the simplest solution (AI that remembers dates) solves billion-dollar problems.

What's next for ConractIQ

Built With

  • badge
  • button
  • card
  • flask
  • languages-typescript-frontend-and-backend-python-whatsapp-bot-(flask)-sql-database-schema-(postgresql)-frontend-react-18-ui-library-vite-build-tool-and-dev-server-react-router-client-side-routing-tailwind-css-styling-shadcn/ui-ui-component-library-(dialog
  • ngrok
  • node.js
  • postgresq
  • python
  • react
  • table
Share this project:

Updates