The problem
A medical bill can be difficult to understand when it arrives with unfamiliar billing codes, unclear line items, and deadlines. People need a way to inspect a bill and prepare informed questions without being told that an AI has made a definitive medical or legal finding.
What MediAudit AI does
MediAudit AI turns a photo or pasted text from a medical bill into a plain-language review. It translates billing terms, flags patterns that may merit follow-up, gives an estimated reference range with confidence and source context, and prepares a checklist and call script. It also offers simpler wording, Spanish output, and links to assistance resources.
How it works
The existing prototype uses a multi-stage pipeline: Gemini 3 Flash reads the bill and masks patient identifiers; retrieval checks charges against a curated reference set; a reasoning step prepares the explanation and next steps; and a second AI pass reviews each flag. The app is built with Next.js, React, and TypeScript and deployed on Vercel.
Safety and limits
MediAudit is an informational aid, not a medical or legal service. Its price ranges are estimates, not a live price database; a flag is a prompt to ask the provider for clarification, not a conclusion that a charge is illegal. The user stays in control of what to share and what action to take.
Prior work and timeline
This is an existing prototype, not a new build for DSH Hacks V2. The public repository history begins September 15, 2026, and the source Devpost page was created September 16. The project was previously submitted to TechCommons Hacks V2 and InfinityX Global Hackathon 2K26, where it received First Runner-Up. This copy preserves that history and does not claim event-specific development.
Built With
- 3
- claude-code
- gemini
- next.js
- typescript
- vercel
Log in or sign up for Devpost to join the conversation.