I study Reinforcement Learning (RL) with the goal of creating general purpose robots that can learn from raw sensorimotor data. I believe that the ability to discover representations that permit decision making over various timescales is a hallmark of intelligence and will allow AI agents to solve hard problems in harsh environments. As a result, I am currently working on solving the skill discovery problem - how can agents automatically break down complex problems into simpler sub-problems by simply interacting with their environment?
I graduated from Harvey Mudd College in 2016, where I was part of the Lab for Autonomous and Intelligent Robotics (LAIR) and was advised by Professor Chris Clark. I then worked at Apple in Cupertino, CA for 2 years as part of the Multitouch Algorithms team under the leadership of Wayne Westerman.
- Dec 2019: Our paper Option Discovery using Deep Skill Chaining has been accepted for ICLR 2020! Can’t wait to go to Ethiopia!
- Dec 2019: I presented a poster on skill discovery at the Deep Reinforcement Learning Workshop at NeurIPS in Vancover, Canada
- Oct 2019: I proposed my research comps on Skill Discovery for Long Horizon Problems. Thanks to my committee members George Konidaris, Michael Littman and Stefanie Tellex for their valuable feedback. Looking forward to making some progress on this problem and defending next semester!
Replication of a Unified Bellman Optimality Principle Combining Reward Maximization and Empowerment
Akhil Bagaria, Seungchan Kim, Alessio Mazzetto, Rafael Rodriguez-Sanchez
Accepted, NeurIPS 2019 Replication Challenge