My main research interest is trying to figure out how to learn good unsupervised world models and how to use them for vastly speeding up reinforcement learning. I have a fondness for probabilistic and bayesian modelling, models that learn by trying to predict the future and models that can be trained end-to-end.


  • Unsupervised Learning
  • Probabilistic Modelling
  • Graph Neural Networks
  • Model Based Reinforcement Learning
  • Generative Models