I am a Researcher at Microsoft Research Cambridge. I focus on Reinforcement Learning (RL), particularly exploration, as applied to both regular MDPs and multi-agent systems.
My research goals revolve around improving sample efficiency, where I believe that Bayesian uncertainty estimates are a promising way to drive exploration. I am also interested in tackling multi-agent systems with modifications of existing RL techniques. I believe that applications in computer games are one of the best drivers for progress in these areas.
Previously, I was a post-doc in Shimon Whiteson’s group at the University of Oxford, where I worked on RL, with emphasis on policy gradient methods. Prior to that, I did a PhD at UCL on RL with linear transition models.