I build and evaluate LLM systems that turn unstructured text into structured causal knowledge. I’m the founder of Causena, where we turn interviews, debriefs and reports into source-cited causal graphs that show what drives real-world outcomes. Before Causena, I was a Postdoctoral Research Fellow at Harvard Medical School / Massachusetts General Hospital in the MJ Lab, where I built benchmarks and extraction pipelines for LLM-generated causal graphs. Before that, I was an AI Resident at Google X, using LLMs to automate causal modeling. I hold a PhD in Industrial and Systems Engineering and an MEng in Computer Science from Virginia Tech.

Selected work

Causal graph benchmark

Reviewed 40+ similarity metrics and validated 9 on 2,000 LLM-generated causal graphs.

Paper · Code

LLMs for causal loop diagrams

Curated prompting turns text into expert-quality causal loop diagrams, cutting analysis time 10x.

Paper · Talk

Causena

Processed 1,000 healthcare transcripts into source-cited causal graphs; adopted by our first customer.

Website

I’m always happy to talk about LLM evaluation, information extraction, and causal AI for messy qualitative data. Reach me at georgialiu63@gmail.com.