Juan Carlos Perdomo

Assistant Professor, NYU CS & Data Science (Starting Fall 2026)

My research develops foundations of machine learning for the social world. I study the feedback loops that arise between people and algorithms, and develop methods with provable guarantees in these dynamic, social contexts.

I earned my PhD in EECS from UC Berkeley, and my BA in CS and Math from Harvard. I was previously a CRCS postdoctoral fellow at Harvard and am spending this year as a postdoc at MIT.

Publications Google Scholar Bio CV

Research Overview

I organize my work around a mix of both theoretical and empirically driven directions. These include,

Performative Prediction: When we use predictions to inform decisions about people, these predictions don't just change the future: they actively shape it. Traffic predictions shape traffic patterns and economic forecasts influence market prices. Performative prediction is a learning theoretic framework we developed to understand learning in these settings.

Prediction in Resource Allocation: Faced with pressure to modernize, large bureaucracies -- such as education and labor departments -- are increasingly using statistical tools to target their interventions efficiently. How do we prove that our prediction algorithms lead to efficient allocations? When is a prediction system "good enough" in these settings? My work in this area changed how the Wisconsin Department of Public Instruction assigned interventions to over 200k students each year.

More generally, I am also interested in learning theory, optimization, and sequential decision making.

People

Interns Hosted: Unai Fischer-Abaigar

Selected Awards & News Coverage

Outstanding Paper Award at ICML 2025 (awarded to 6 papers, out of 10k+ submissions). [Coverage]
NSF Graduate Research Fellowship
Hoopes Prize Winner
Under 17, 19, and under 21 Sailing World Champion for Puerto Rico in Laser Radial (ILCA 6)

Contact

j.perdomo.silva[at]nyu.edu