Congratulations to Kaustubh Sridhar on receiving the Charles Hallac and Sarah Keil Wolf Award for Best Dissertation from the Department of Electrical and Systems Engineering.
Supervised by Prof. Insup Lee, his dissertation, “Training Adaptive and Sample-Efficient Autonomous Agents,” explores how AI agents can generalize beyond their training data and adapt to new tasks and environments using minimal additional data. This work advances data-efficient methods for imitation and reinforcement learning, introducing systems such as REGENT and RICL that enable adaptable, generalist AI agents for embodied tasks and robots.
Since completing his PhD in May 2025, Sridhar has joined Google DeepMind as a Research Scientist. There, he is building open-ended world models and general embodied agents—directly extending his PhD research to push the boundaries of adaptive, real-world AI systems, including contributions to projects like SIMA.
The full dissertation is available via ProQuest, and a PDF version can be accessed here.
Congratulations to Kaustubh on this well-deserved recognition.