Claude studied deep reinforcement learning and graph neural networks in his honours year and is now working on multi-agent RL.