home page showing lost or found items
user profiles
search for a lost item
where matchers are placed
post lost item

Inspiration

What it does

How we built it

Challenges we ran into🧩 Project Story — Findr: Smart Lost & Found App

💡 About the Project

Findr is an AI-powered lost and found mobile application designed to help users quickly reunite with their belongings. The app allows users to upload photos or text descriptions of lost or found items. Using Gemini’s multimodal AI, it analyzes both images and text to automatically match related posts. The goal was to make lost-and-found systems faster, smarter, and more reliable—replacing manual searches with intelligent automation.

🌱 Inspiration

The idea was inspired by the frustration of losing items around campus and relying on outdated notice boards or WhatsApp groups. We wanted to leverage AI to make the process effortless—so people could simply “snap it, match it, and get it back.”

🛠️ How We Built It

Frontend: Built with React Native for cross-platform mobile support.

Backend: Implemented as serverless functions on Vercel, ensuring scalability and zero maintenance overhead.

Database & Storage: Powered by Supabase, used for authentication, real-time updates, and secure image storage.

AI Layer: Integrated the Gemini API for vision and language understanding. The AI generates item captions and computes similarity between lost and found reports based on visual and textual embeddings.

The AI’s similarity function can be abstractly represented as:

𝑆 ( 𝐴 , 𝐵

)

𝛼 ⋅ sim image ( 𝐴 , 𝐵 ) + ( 1 − 𝛼 ) ⋅ sim text ( 𝐴 , 𝐵 ) S(A,B)=α⋅sim image

(A,B)+(1−α)⋅sim text

(A,B)

where 𝑆 ( 𝐴 , 𝐵 ) S(A,B) is the combined similarity score, and 𝛼 α balances the image and text contributions.

🧠 What We Learned

How to integrate multimodal AI (text + image) with a serverless architecture.

The importance of clean data and consistent image quality for reliable AI results.

How to use Supabase Auth and Storage efficiently in a mobile environment.

Designing an intuitive user experience that simplifies complex AI processes.

⚙️ Challenges

Handling inconsistent lighting and angles in user-uploaded photos.

Managing Gemini API limits while maintaining real-time response speeds.

Balancing accuracy vs. cost in AI inference calls.

Deploying serverless functions with proper security and environment isolation.

🚀 Outcome

Findr successfully demonstrates how AI and serverless computing can simplify everyday problems. It transforms a common pain point—losing items—into a seamless, intelligent experience that saves time and effort.