ClaimCure

Front page when the user first enter the website
Scroll down or click on the about us button will take the user here
Example of a training data
Loading
Uploading
Scanned

Inspiration

The Burden of Medical Debt "Approximately 14 million people (6% of adults) in the U.S. owe over $1,000 in medical debt and about 3 million people (1% of adults) owe medical debt of more than $10,000.” – ILR, Cornell University The Cost of Overcharging "The 100 most expensive U.S. hospitals charge from $1,129 to $1,808 for every $100 of their costs. Nationally, U.S. hospitals average $417 for every $100 of their costs, a markup that has more than doubled over the past 20 years.” – National Nurses United

What it does

Take a photo or scan of your hospital bill and upload it to ClaimCure. Our system uses AI to parse hospital bills, check for errors or inflated costs, and compare the charges against standard rates. It drafts an official email to the hospital’s billing department, pointing out potential overcharges and asking for an adjustment.

How we built it

Frontend: Built with React, Deployed on Vercel Backend: Python + FastAPI framework for handling AI logic and billing analysis. Hosted on Render in Docker containers Document Analysis: Uses Google Document AI (or OCR libraries) to parse uploaded hospital bills. AI-generated emails are sent to hospital with documents

Challenges we ran into

Not enough data about medical bills images, DocumentAI limitations and API unclarity.

Accomplishments that we're proud of

Despite a lack of data, we were able to train a model to sufficiently extract data from medical bills and then turn that extracted data into claims about how fair the prices are.

What we learned

We learned to use cloud services embedded within our application and utilize their capabilities at no cost to our local machines capabilities.

What's next for ClaimCure

We will improve upon our system for detecting whether or not the price for a procedure is over charged by deploying a machine learning anomaly detection model to help identify outliers and how big of an outlier they may be.