AutoCurve
Assistant
A hybrid car-valuation engine that pairs statistical market pricing with an AI assistant capable of judging a vehicle's physical condition from photos.
Overview
Pricing a used car usually means trusting a listing price or a gut feeling. AutoCurve combines two signals instead: a statistical model built from scraped historical market data, and a vision-language model that inspects the actual car in the photos you upload — so the estimate reflects both the market and the machine in front of you.
What I built
Statistical pricing model
Designed a model that estimates current fair-value ranges from scraped historical datasets, turning noisy listing data into a defensible price band.
Vision-based condition analysis
Integrated a vision-language model that analyzes user-uploaded vehicle photos and returns structured condition flags — dents, wear, mismatched panels — that adjust the valuation.
Python backend & REST API
Built the backend and the REST API layer that connects the valuation engine to the AI evaluation module, keeping the two systems cleanly decoupled under hackathon time pressure.