Inspiration
Took the inspiration from a real-life case study : A NYT case study revealed patients overpay millions due to unreadable medical bills. We built Bill Lens to give everyone a fair, AI-powered lens into their healthcare charges.
https://www.nytimes.com/2026/04/08/health/ai-chatbots-medical-bills-claude-chatgpt.html
What it does
Bill Lens lets users upload medical bills and instantly get a clear, plain-language breakdown flagging errors, overcharges, and confusing line items using AI vision analysis.
How we built it
Used Gemini for research and architecture planning, then built the full end-to-end product with Claude covering UI, backend logic, prompt design, and GROQ-API integration.
Challenges we ran into
No open source vision model available. No cloud provider. All required Local setup.
Accomplishments that we're proud of
- Integrated GROQ-API for the bill understanding and analysis
- Handled Fallbacks with grace
What we learned
Prompt design is everything in medical AI. Small wording changes drastically affect output quality. Vision models still struggle with dense, formatted documents like hospital bills.
What's next for Bill Lens
Prompt Engineering or Fine-tuning of vision language models for medical bill understanding
Log in or sign up for Devpost to join the conversation.