Selected work
A few examples of machine learning systems built for practical, real-world use.
Built and deployed RetinaNet and YOLO models to active construction sites, processing CCTV streams to identify PPE violations and safety hazards in real time.
The work moved from model development into an operational setting, where latency, camera conditions, and reliable alerting mattered as much as benchmark accuracy.
Created an image-based tracking application that uses segmentation and classification to measure consumption on airline meal trays.
The resulting data enables large-scale analysis of passenger preferences and supports more targeted reduction of food waste.
An implementation project exploring the ideas behind Pluribus, Facebook AI Research’s multiplayer poker system.
Built as a practical way to learn reinforcement learning, imperfect-information games, and multi-agent decision-making.
View on GitHub ↗