research

I build LLM-enabled agent systems that extract information from messy, heterogeneous sources and assemble it into large knowledge networks that we can analyze with network science. My work focuses on making these synthesized graphs trustworthy by developing ways to evaluate their correctness and completeness, and then using them to study questions like how scientific ideas spread and how researchers move across fields and institutions. I also work on the systems side of multi-agent AI, including scheduling and serving optimizations that make compound LLM applications more efficient at scale.