Evan R

I’m an AI developer and data practitioner who builds machine learning systems with a strong focus on reinforcement learning and practical, end-to-end application delivery. I combine structured data thinking with hands-on model training and debugging to produce reliable intelligent systems.

Most of my recent momentum comes from training and deploying an AlphaZero-style reinforcement learning pipeline for a full-stack board-game project. I designed and implemented a full RL setup (including a custom game engine), built MCTS with a joint policy/value head, and ran iterative self-play training across parallel workers on Modal.com.

I’m deliberate about training stability and output quality. When I saw real pathologies—broken MCTS priors, value head miscalibration, and rollout bottlenecks—I diagnosed the root causes and adjusted the training workflow to improve model stability and the quality of self-play decisions.

Beyond RL, I also develop applied ML for everyday utility, like a Python-based Spotify song recommendation model built with Scikit-learn and validated using engineered audio features and visualization. My background in operations and high-volume client management sharpened my ability to track complex workflows, maintain data accuracy, and deliver under real constraints—skills I bring directly into building dependable AI systems.