Aidan Sirbu

Howdy!

I'm a PhD student broadly interested in self-supervised learning and generative modelling. I work on generative motor control of embodied agents.

Both halves of my work come down to the same question: what does a system actually have to compute in order to act well? On the embodied side, I build simulated animals faithful enough that theories of motor control can be validated against measured behaviour. On the other, I study models that solve problems by repetition, where one shared update run to arbitrary depth does the work usually handed to search or to scale.

Before I joined Mila, I completed my BSc in Medical Physics at Western University where I worked on projects exploring hyperpolarized magnetic resonance imaging.

In my free time, I am an avid rock climber, hiker, and motorcycle rider.