Aviral Kumar

About

I am a second year PhD student in the Berkeley AI Research (BAIR) Lab at UC Berkeley advised by Prof. Sergey Levine. My research interests lie in the field of Machine Learning and Reinforcement Learning. My research goal is to develop agents that can learn via interaction, reliably in the real world. More concretely, this involves understanding the interaction between deep neural nets, optimization and interactive reinforcement learning algorithms and using these insights to develop better algorithms. Further, I am interested in developing ways to effectively and efficiently use prior, off-policy datasets, including sub-optimal demonstrations while training an RL agent. I previously interned at the Google Brain Team, Toronto during Summer 2017 and Summer 2018 where I was fortunate to work with Geoffrey Hinton, Kevin Swersky and Jimmy Ba on a variety of topics including Hierarchical RL, relational learning and generative modelling.