Ethan Villalovoz

Sacramento, California, United States

Portrait of Ethan Villalovoz

I work on robot learning at Georgia Tech. Most recently, I built and evaluated enterprise AI agent systems at Microsoft. I'm interested in how robots learn from imperfect human feedback, reason under uncertainty, and align their behavior with human intent.

You can reach me at .

BODE-GEN (arXiv 2025)

Bayesian prompt optimization for test-driven code generation, evaluated across 164 HumanEval+ tasks and three code models.

A graph-augmented retrieval system combining semantic search, DBpedia relationships, and source-grounded answers across a 10,000+ article knowledge base.