Kate Saenko

Kate Saenko is a computer scientist whose work helps visual AI adapt across image domains and tests how models explain their decisions. She earned a computer science PhD at MIT in 2009, completed postdoctoral fellowships at UC Berkeley and Harvard, served on the UMass Lowell faculty, and joined Boston University as an assistant professor in 2016. Her research includes Deep CORAL, which aligns deep-network features across domains, and RISE, a method for testing explanations of black-box models. In 2025, she coauthored KiVA, a visual-analogy benchmark for large multimodal models; she is a professor at Boston University and an AI Research Scientist at FAIR Labs, Meta.

Now at: Boston University and Meta
Works: 354
Citations: 54,115
h-index: 82
Badges: Rising
Bridge
Wayfarer

Publication numbers from OpenAlex, as of 22 October 2026.

Timeline

By 2016

MIT doctorate and postdocs

In 2009, she earns a PhD in computer science from MIT. By 2016, she completes postdoctoral fellowships at UC Berkeley and Harvard University.
2010

Adapting visual category models

In 2010, Saenko coauthors Adapting Visual Category Models to New Domains with Brian Kulis, Mario Fritz, and Trevor Darrell. The method learns a transformation that adapts object models to new imaging conditions.
2016

Assistant professor at BU

In 2016, she is an assistant professor of computer science at Boston University, following a faculty appointment at UMass Lowell. Her research covers machine learning and computer vision.
2016

Deep CORAL for domain adaptation

With Baochen Sun, she coauthors Deep CORAL in 2016, extending CORAL with a nonlinear transformation that aligns correlations between deep-network layer activations across source and target domains.
2017

A CVPR prize and DARPA grant

In 2017, Saenko and her students Huijuan Xu and Abir Das win the Most Innovative Solution prize on the IEEE Large Scale Activity Recognition Challenge at CVPR. Saenko and Trevor Darrell receive a four-year, $7.55 million DARPA grant to develop a tool that explains AI decision processes; $800,000 is allotted to her lab.
2018

RISE for black-box explanations

In 2018, she and collaborators present RISE: Randomized Input Sampling for Explanation of Black-box Models at BMVC. The method highlights image regions that a neural network uses, then checks whether masking those regions changes the network's prediction.
2025

KiVA tests visual analogies

In 2025, Saenko coauthors KiVA, a set of 4,300 visual transformations that tests whether large multimodal models can infer visual analogies and apply them to new objects. The study compares model performance with children and adults.
As of October 2026

Professor and Meta AI researcher

As of October 2026, Saenko is a Professor of Computer Science at Boston University and an AI Research Scientist at FAIR Labs, Meta. She directs BU's Computer Vision Learning Group and is founder and co-director of the Artificial Intelligence Research Initiative.