FoundIt

• 

  Lost Item Option

• 

  First Page

• 

  Found item option

• 

  AI matches the lost and found items and suggests potential matches

Inspiration

FounIt is an AI-powered lost and found platform designed for university campuses. It allows users to post lost and found items without needing a login, using AI to match descriptions and images. This has always been one of the main problem in college when people used to spam multiple whatsapp groups when they find earpods,mobiles,headset or any other easyily misplacable item.I get it,as humans we tend to forget and loose things under pressure :) why is why I created FoundIt. It allows people to post lost and found items with a title,suitable description if any,upload pictures and contact details.It is the go-to place for students! Instead of scrolling through hundreds of messages, our AI engine automatically compares "Found" reports against existing "Lost" entries using both image recognition and semantic text analysis.

What it does

Foundit allows users to post images of lost and found items around their university campus and lets AI matches items by scanning hundreds.Users can report an item in under 30 seconds. No login is required, ensuring that someone who finds a lost item isn't deterred by a long sign-up process.It replaces fragmented social media groups with a single, searchable database dedicated specifically to campus recovery.Instead of manual searching, FoundIt uses a vector-based matching engine. It bridges the gap between a "blue water bottle" and a "Hydroflask found in the library" automatically.

How we built it

Built a high-performance backend using FastAPI and SQLAlchemy and integrated Sentence Transformers to convert item descriptions into vector embeddings, allowing us to calculate mathematical similarity between "Lost" and "Found" entries.The frontend was developed with Next.js and Tailwind CSS. We used Framer Motion to create a fluid, mobile-first experience that feels like a native app.Used Pillow for server-side image handling to ensure all user uploads are optimized and consistent before being displayed in the feed.

Challenges we ran into

1. Designing the logic to trigger a match search immediately upon a new post without causing latency in the UI.
2. Syncing the asynchronous nature of AI processing with the real-time updates on the Next.js frontend.

Accomplishments that we're proud of

1. Successfully implementing a "No-Login" flow that still feels secure and organized.
2. Moving beyond simple keyword matching; our system understands that a "laptop charger" and a "MacBook power brick" are likely the same thing.

What we learned

Learnt how to effectively deploy machine learning models within a FastAPI environment to provide real-time utility. The importance of choosing the right tools(in this case SQLite) for rapid prototyping during a hackathon.Understanding that in a lost-and-found scenario, speed and ease of use are more important to the user than complex account features.

What's next for FoundIt

1. Implementing WebSockets so that users get an instant alert the moment a potential match is posted. 2. Moving beyond text embeddings to incorporate Computer Vision allowing AI to compare the actual pixels of two uploaded photos for a match.
2. Creating a "Verified Student" toggle using University SSO to add a layer of trust for high-value items like laptops or wallets.

Built With

• fastapi
• framermotion
• nextjs
• pillow
• sqlite
• tailwindcss
• transformers