Research Scientist

DeepMind, London

anna -at- harutyunyan -dot- net

Keywords:  reinforcement learning, temporal abstraction, off-policy learning



My work is towards designing principled reinforcement learning algorithms that are able to leverage structure, discover abstractions, and generalize across environments. I would also like to get rid of the primitive time step one day. Before joining DeepMind, I completed my PhD in 2017 at the AI lab of VU Brussel, working with Prof. Ann Nowe and Peter Vrancx on eclectic reinforcement learning algorithms. I was also involved in an assistive assistive exoskeleton project, where I was responsible for providing the high-level control, and doing a lot of prediction. Before this, I received my Masters degree at Oregon State University with the amazing Prof. Cora Borradaile, working on max flow algorithms in planar graphs.

My (likely outdated) CV can be found here, if you are into that kind of a thing.