Current Work
What I'm currently focused on:
AI Engineer — Robyn Dawes Institute
Building an AI-native pipeline to extract structured data from 213+ development economics RCTs. Working on PDF table parsing (TEI XML to markdown), LLM-based field extraction, and a human-in-the-loop validation tool — part of a full-stack research platform (FastAPI, Next.js, Supabase, Docker).
Research Assistant — UC Berkeley Haas
Continuing research on confidence calibration in large language models under Prof. Don Moore. Developing computationally efficient methods for quantifying LLM uncertainty and flagging over-confident behavior for downstream fine-tuning.