Writing

Technical articles on engineering problems I've worked through: infrastructure, observability, systems design, and the reasoning behind architectural decisions.

• July 22, 2026Reward Design Is the Hard Part: Lessons From Training Autonomous Surface Vessels With Reinforcement Learning

  Reinforcement learning gets described as if the algorithm is the interesting part. A Leidos competition to control unmanned surface vessels shows the real difficulty is in observation spaces, action spaces, and reward design.

• June 25, 2026Building Log Patrol: AI-Assisted Log Monitoring That Actually Reduces Alert Fatigue

  How I built a three-layer monitoring system using deterministic rules, Drain3 template mining, and a local LLM sentiment gate to eliminate noise from Loki log streams.