ToyotaMatch - Hackathon Submission

Inspiration

Finding the perfect Toyota or Lexus shouldn't require navigating complex filters or getting "no results found." We wanted to create an AI-powered car marketplace that understands natural language, never gives up, and lets users explore vehicles in immersive 3D.

What it does

ToyotaMatch is an intelligent car marketplace that combines AI-powered natural language search with real-time inventory data. Users can search using plain English like "sports car low mileage" or "under $25k near Austin," and the system intelligently finds matches even when inventory is limited. Key features include:

AI-Powered Search: GPT-4o understands complex queries and converts them to precise API calls
Smart Constraint Relaxation: Never see "no results" --- the system automatically widens parameters while showing what changed
Interactive 3D Model Viewer: Explore car interiors in stunning 3D with drag-to-rotate, zoom, and fullscreen modes
Comparable Models Discovery: Every listing shows intelligent alternatives with smart deduplication
Full Explainability: See exactly why each result matched with "Match DNA" breakdowns

How we built it

Built with Next.js 16, React 19, and TypeScript. We parse the user input and dynamically determine the best possible listings.

Challenges we ran into

The biggest challenge was implementing the 3D model viewer. We had to:

Convert and optimize GLB files for web performance
Handle camera controls, lighting, and responsive rendering
Ensure smooth interactions across different devices
Map car models to their corresponding 3D files dynamically
Balance visual quality with load times
Our live website could not handle the size

Accomplishments that we're proud of

Zero empty results: Our relaxation ladder ensures users always see options, even with strict filters
3D integration: Successfully integrated interactive 3D models for 10+ Toyota/Lexus vehicles
AI accuracy: GPT-4o parsing handles complex, conversational queries with high accuracy
Performance: Smart caching and pool management reduce API calls by 80%+ when browsing comparables
Transparency: Every decision is explainable users see exactly why results matched

What we learned

We learned that combining AI with real-time inventory data requires careful prompt engineering and robust error handling. The 3D model integration taught us about WebGL optimization and the importance of progressive loading. Most importantly, we discovered that users value transparency showing what constraints were relaxed builds trust even when exact matches aren't found.