On my AI Chess Coach project, I designed an architecture that separates deterministic chess reasoning from LLM-based coaching feedback. I integrated Stockfish with heuristic agents and built a PGN review engine for move-by-move replay.
I built a phishing detection MLOps pipeline that achieved 92% accuracy across 50,000+ records and served real-time predictions. I tracked model versions with MLflow and containerized training and inference services with Docker.
I also contributed to Ray Data by resolving a naming ambiguity across four MapTransformFn implementations and updating the corresponding unit tests. My work spans machine learning, generative AI, and open-source contributions.
