Found @ OSU

Claiming Sample Item
Directions to find claimed Item
Main Dashboard
Posting Lost Item

Found @ OSU | Builder Track

Inspiration

Lost items on college campuses are a daily frustration. Anyone can claim "that's mine!" but how do you really know? We wanted to solve this trust problem with technology that actually works. Traditional lost and found systems rely on manual verification or simple descriptions that anyone could guess. We envisioned an AI-powered platform that could intelligently verify ownership without requiring physical presence.

What it does

Found @ OSU is a smart lost-and-found platform where finders upload photos of items they discover. Our AI automatically generates unique security questions based on the item's details that only the true owner would know. When someone tries to claim an item, they must answer these questions. This prevents fraudulent claims while being forgiving of slight wording differences.

How we built it

Frontend: React with TypeScript, Tailwind CSS, and shadcn-ui components for a modern, responsive interface
Backend: Lovable Cloud (Supabase) for database, authentication, and serverless edge functions
AI Integration: Lovable AI Gateway for image analysis and question generation using Google Gemini models
Mapping: Mapbox GL for interactive location tagging and display
Search: Fuse.js for fuzzy search across lost items

Challenges we ran into

The biggest challenge was implementing reliable semantic verification. Initially, we tried simple text comparison, but it failed when users phrased answers differently. Another challenge was ensuring the AI-generated questions that were specific enough to prevent false claims but flexible enough for legitimate owners to answer.

Accomplishments that we're proud of

We built a fully functional platform in record time that genuinely solves a real problem. The semantic similarity verification is particularly impressive. It can understand that "blue Nike backpack" and "navy swoosh bag" refer to the same thing. The AI question generation creates truly unique, context-aware security questions that would be nearly impossible for someone who didn't own the item to answer correctly.

What we learned

We discovered the power of combining multiple AI models for different tasks - using vision models for image analysis and language models for verification creates a robust system. We also learned that semantic similarity is far superior to exact matching for real-world applications where human language varies naturally.

What's next for Found @ OSU

We plan to expand to multiple universities, add push notifications for new item matches, implement a reputation system for frequent finders, and potentially integrate with campus security systems. We're also exploring using the platform's data to identify high-loss areas on campus to help prevent future losses.