BeCarful

Inspiration

Car accidents are already stressful, and dealing with insurance afterward can make things even worse. Drivers have to document damage, understand a complicated policy, estimate repair costs, and figure out what to submit for a claim. We built BeCarful to make that process simpler and keep everything in one place.

What it does

BeCarful helps drivers document and manage an accident from start to finish. Users can take or upload photos of their car, upload their insurance policy, and use an AI assistant to understand their coverage and damage. The app then creates a summary with estimated repair costs, next steps, and a pre-filled insurance claim document that the user can review and download.

How we built it

We built BeCarful with Next.js and TypeScript, with MongoDB for application data and AWS S3 for storing photos and insurance documents. We use Gemini to analyze vehicle damage, understand insurance policies, answer questions, and help generate claim information. We also use Three.js to give users an interactive 3D view of their vehicle and visually highlight damaged areas.

Challenges we ran into

One of our biggest challenges was connecting AI results to the rest of the application in a reliable way. Instead of letting the AI directly modify the 3D model, we made it return structured vehicle-part information that our app can validate and map to the correct parts of the car. We also had to think carefully about handling sensitive documents and keeping information separated between different vehicles.

Accomplishments that we're proud of

We're proud that BeCarful is more than just an AI chatbot. Photos, insurance documents, damage analysis, the 3D vehicle, chat, and claim preparation all work together as one workflow. We especially like that a user can go from documenting an accident to having a clear list of what to do next and a claim document ready to review.

What we learned

We learned that using AI effectively is less about adding it everywhere and more about deciding where it is actually useful. Gemini works well for understanding unstructured information like photos, policies, and user questions, while important workflow decisions are better handled by predictable application logic.

What's next for BeCarful

Next, we want to improve damage and repair-cost estimates, support more insurance providers and claim forms, and make the 3D damage visualization more detailed. Longer term, we'd like BeCarful to guide users through the entire claim process—from the moment an accident happens until the claim is resolved.