Building Reward Signals for LLM Agents
Lessons from designing evaluation frameworks that actually measure what matters — not just what's easy to measure.
Feb 2025 Draft
Writing
Thinking in public.
Notes on building AI systems, the math underneath, and the human side of engineering work.
Jan 2025 Draft
From Mathematician to ML Engineer
How a Ph.D. in pure math turned into a career building AI systems at scale, and why the transition was less linear than it sounds.
Jan 2025 Draft
The Case for Auto-Raters in Agent Evaluation
Why human evaluation doesn't scale for agentic systems, and how to build automated evaluation that you can actually trust.
✦ More posts coming soon. In the meantime, you can find my thinking scattered across commit messages and design docs.