Inspiration

Hospitals overcharge and get away with it. This is an uncomfortable fact that you don't notice in your daily life. But when my dad was hit by a car, my family was faced with it. 1,000 dollars for medical implements, like gauze? Hundreds more for simple pills, like ibuprofen? It felt wrong, and LLMs alone can't coach you through the painful call.

What it does

We build your case, and coach you through it. Instead of browsing the web from uncertain, disparate sources for prices, we get our fair pricing data directly from the US CMS. Then, we write you a dispute letter, and when you call the hospital to dispute your bill, our chatbot listens & coaches you through the conversation.

How we built it

We used a React + TypeScript frontend, and a Python FastAPI backend. To parse the bill accurately, we use Claude Vision. We use sentence-transformers + FAISS for real-time CPT code matching against the Medicare database. For fast or complex reasoning, we use either Claude Haiku or Claude Sonnet.

Challenges we ran into

Ideating was hard, especially when considering Claude's strengths and creating something with real impact. But once we arrived at this idea, we knew it was the one.

Accomplishments that we're proud of

The live coach that differentiates between the company rep and overcharged patient was a point of pride for us.

What we learned

Don't start design too early! Build up your backend and then tailor the frontend to your purposes.

What's next for ClearBill

YC, Clearly.

Built With

  • claude
  • faiss:
  • fastapi
  • haiku
  • python
  • react-+-typescript-frontend
  • sentence-transformers
  • sonnet
  • vision:
Share this project:

Updates