Judy Hoffman
Judy Hoffman completed a Berkeley doctorate in 2016 and held postdoctoral roles at Stanford and Berkeley. In 2018 she co-authored CyCADA, a method combining image-space and feature-space alignment for domain adaptation. Her career includes FAIR and Georgia Tech work, and she moved to UC Irvine in January 2026. Her research focuses on data-efficient adaptation, reduced computation, and tools to understand model errors.
- Works
- 114
- Citations
- 19,483
- h-index
- 46
Publication numbers from OpenAlex, as of 14 October 2026.
Timeline
-
2016
A doctorate and postdoctoral research
Hoffman earned a Berkeley EECS doctorate in 2016 with Trevor Darrell. She then held postdoctoral roles at Stanford and Berkeley.
-
2018
Cycle-consistent domain adaptation
Hoffman co-authored CyCADA at ICML 2018. The method adapts between domains using image-space generation and alignment of latent representations. It uses task guidance and semantic consistency to reduce changes in meaning during adaptation. The authors report visual recognition and road-scene experiments.
-
Earlier career
Research on reliable visual learning
She conducted research at FAIR and held an associate professorship at Georgia Tech before Irvine. Her research focuses on data-efficient adaptation, reduced computing time, and benchmarks for understanding and reducing model errors.
-
2026
A move to Irvine
She moved to UC Irvine in January 2026. She is a computer science associate professor and directs its Center for Machine Learning. She also holds an adjunct role at Georgia Tech.
Sources and credits
Photo: jhoffman.github.io; creator not stated. License: Not stated by source. Source: jhoffman.github.io. Cropped to a square and resized. Ask us to remove or replace this photo.