I’m an applied AI scientist working on agent systems and LLM reliability. My background spans [insert: brief, e.g. “a PhD in NLP, DARPA-funded research in machine reading, and several years of industry work on conversational AI systems”].
I currently spend my time on the methodology of building AI agents that actually work — evaluation, failure analysis, and the engineering decisions that sit around the LLM (the “harness”). Most recently, this has meant studying how harness design choices affect reliability in bounded-corpus deep research agents.
This site is where I publish writing on applied AI, agent reliability, and the craft of shipping systems under real-world constraints.