One-Man-Crew

• 

  project work flow

• 

  similarity_ranking

Inspiration

This project was inspired by personal experience. I have lost valuable items before, some financially valuable and others emotionally important, such as a graduation gift from my mom. I realized how difficult it can be for students to recover items even when they are found, and I wanted to build something that could help reduce that frustration.

What I Built

I built an AI assisted Lost and Found system that helps admins match lost item inquiries with found inventory while keeping humans in control. The system uses image embeddings, text embeddings, and similarity ranking to surface the top five most relevant matches. Instead of automatically resolving cases, admins review results, ask clarification questions, and confirm ownership manually.

What I Learned

Through this project, I learned the importance of human in the loop design. AI works best when it supports human judgment rather than replacing it.

Challenges

The main challenges were balancing automation with trust, designing clear workflows, and handling incomplete data while keeping the system reliable.

Impact

This project can benefit many students by making lost item recovery faster, safer, and more reliable. By combining AI assisted matching with human verification, the system reduces false matches, prevents fraud, and helps ensure that valuable and sentimental items are returned to their rightful owners.

Built With

• cosinesimilarity
• css
• flask
• geminiapi
• html
• imageembeddings
• javascript
• json
• pandas
• python
• react
• smtp
• textembeddings