I am working on task-aligned and value-aware model learning for reinforcement learning and control. My research focuses on agents learning world models which are correct where it matters, meaning they can adapt their losses to the task at hand.

I am also interested in the learning dynamics of Deep Learning agents, using mathematical tools from optimization, dynamical systems and learning theory to understand what RL agents learn (and fail to learn).

In my spare time, I volunteer at Queer in AI, an affinity group that is working to protect queer scientists working on and users of AI systems.