I'm a Senior Machine Learning Engineer at Bloomberg, working on LLM-powered tools for information extraction and analyst workflows.
My current interests are practical AI agents, evaluation, reinforcement learning, grounding, tool use, and the engineering systems that make LLM workflows reliable enough for real users.
Outside work, I write and build around those themes:
- ryanlee.ai — notes on AI agents, LLMs, web extraction, and practical automation
- Agents in Practice — a weekly field note on agent research, benchmarks, and lessons from using agents in real workflows
- Hermes Agent as My Day-to-Day AI Interface — a concrete example of the kind of human/agent workflow infrastructure I'm exploring
Previously, I studied mathematics at Princeton, with minors in computer science and machine learning.
- AI agents that do useful work beyond demos
- LLM evaluation, reinforcement learning, reliability, and failure analysis
- Grounded extraction from messy documents and web data
- Research-adjacent engineering: tools, experiments, benchmarks, and systems for understanding model behavior
- Website: ryanlee.ai
- LinkedIn: linkedin.com/in/seungjaeryanlee