Diego Garcia-Olano

Diego Garcia-Olano is a Senior Research Scientist at Meta, who has worked across Responsible AI, Meta Superintelligence Labs Trust, and ML Innovation for Products on safety alignment, LLM interpretability, and user journey optimization. His research spans interpretability and evaluating and mitigating risks–hallucination, memorization, PII and IP exposure–in LLMs, with publications at NeurIPS, ICLR, EMNLP, ACL, etc. He was a core contributor to Meta’s Llama 3 and 4 releases, co-organizes the Unlearning and Model Editing (U&ME) workshop, and serves on Humane Intelligence’s advisory group. He holds a PhD from UT Austin on interpretable multimodal representation learning for health/social good. As a Fulbright Specialist, he helped build models to predict gender bias in judicial proceedings for an Argentine human rights center; held research internships at Google Research and IBM, and was a Data Science for Social Good fellow at the University of Chicago, working with SEDESOL to improve social-services distribution and alleviate poverty in Mexico.