Built for the Springineering Challenge 2026 at the University of Oulu: a tool that searches security-camera footage from plain-English questions and raises real-time alerts when objects of interest appear. It was designed to deploy online, and demonstrated running fully offline for the challenge.
I designed the computer-vision pipeline that detects, tracks and indexes objects, so footage can be searched by meaning instead of keywords, and built the full application around it: a Python (FastAPI) backend and a React web interface with a built-in AI assistant for asking questions about the footage.
There is no public repository for this project. Code available on request.