Damage assessment for salvage and auction cars: upload photos, get damaged parts, OEM codes, prices, and the real cost of buying and fixing the car.
Inspiration
I bought a damaged car with 10 minutes to decide, estimated the repair by eye, and then spent weeks hunting for parts and OEM numbers. Carbender is the tool I wish I’d had, both before bidding and after buying.
What it does
Upload photos of a damaged car and get a report with:
Damaged parts and severity, shown on an interactive exploded-car damage map OEM part numbers with confidence levels, plus new, aftermarket, and used prices side by side Likely related parts (brackets, trims, hinges, tow hook covers) Labour estimates and a repair-vs-total-loss verdict Auction title and seller status checks, and NHTSA recall lookups US import and customs calculators, a flipping calculator, and a repair planner Country-specific warnings (ADAS calibration, title rules, EU conversion issues) PDF export
Works for both US (DOT/SAE, USD) and European (ECE, EUR) vehicles.
How I built it
Next.js, TypeScript, PostgreSQL (Prisma), Stripe, Inngest for background jobs, Gemini and Cloudflare R2. The OEM matching engine is my own code, backed by a database of 1M+ parts records covering 60+ makes.
Challenges
Left/right confusion in vision models, solved with geometric validation instead of trusting the model’s text. Generation and facelift ambiguity in parts catalogs, solved with chassis-code and year-range validation. Wrong OEM codes shown with confidence are worse than none, so every code has a confidence tier and unverified ones are hidden.
What I learned
Domain logic around the AI matters more than the AI itself. Parts ecosystems are regional even when cars are global, and honest uncertainty builds more trust than false precision.
What’s next
Secondary (hidden) damage prediction, a browser extension for Copart/IAAI, and body shop team accounts.
Live: carbender.com
Built With
- antigravity
- cloudflare-r2
- gemini-api
- google-genai
- google-search-grounding
- inngest
- neon.tech
- next.js
- node.js
- postgresql
- prisma
- react
- sharp
- stripe
- supabase
- tailwind-css
- typescript



Log in or sign up for Devpost to join the conversation.