Csaba Szepesvari

Csaba Szepesvari is a Professor in the Department of Computing Science at the University of Alberta and Canada CIFAR AI Chair at Amii. His research focuses on developing efficient learning algorithms for sequential decision making problems, with particular emphasis on reinforcement learning theory, online learning, and bandit algorithms.

Research interests include:

• Foundations of reinforcement learning and online decision making
• Convergence properties of learning algorithms
• Bandit problems and exploration-exploitation tradeoffs
• Function approximation in machine learning

His publications demonstrate consistent theoretical contributions to understanding algorithmic convergence, complexity, and efficiency in reinforcement learning. Recent work explores LLM uncertainty estimation, policy gradient methods, and offline-to-online learning transitions.

Leadership roles include:

• Foundations team lead at DeepMind
• Organizer of RL Theory Virtual seminar series

He mentors graduate students in theoretical machine learning through the University of Alberta and Amii research programs.