Last updated August 4, 2023.

I currently work as a Research Scientist on the Robotics@Google team, which is part of Google DeepMind. The views expressed on this blog are personal opinions, and do not reflect the opinions of my employer.

My main research interest is deep reinforcement learning and imitation learning, as applied to robotics. In general, I like thinking about what’s necessary to make robot learning work for real-world problems. Much of my work has been based on how to most efficiently use real robot time, how to leverage external data sources that aren’t bottlenecked by real robot hardware, and the challenges of longterm agent-centric data collection.

I graduated from the UC Berkeley Computer Science program in 2016, doing undergraduate research in the Berkeley Artificial Intelligence Research (BAIR) Lab, mentored by Pieter Abbeel and working most closely with John Schulman. I was an Honorable Mention for the NSF Graduate Research Fellowship Program.

During my last year of undergrad, I was very unsure whether I wanted to keep doing research, and eventually decided not to apply to PhD programs (a blog post I wrote around that time can be seen here). However, I have an offer from the Google AI Residency Program, and decided I would try this research thing for one more year. I ended up liking ML research and have been at Google ever since. A comparison of industry research versus academia PhD can be seen here.

As someone who started in math, I’ve always had a soft spot for theoretical computer science. Outside of machine learning, I enjoy complexity theory, theoretical cryptography, and mathematical logic, although I’m not as good at those subjects as I used to be.

This site doubles as my personal blog. Posts range from discussion about machine learning to observations of my life and stupid jokes. I post whenever I have the time and motivation to do so, meaning not very often. Historically I average about one post per month.

In my free time, I primarily play card games and video games. My main games are Magic: the Gathering and Dominion. I’ve never been very good at MTG, but at my peak I was top 20 worldwide at Dominion. (These days it’s more like top 500.)

I’m a lapsed fan of My Little Pony: Friendship is Magic. “Lapsed” because, well, it’s complicated, I have a post covering why. I keep up with little of the fandom these days, but much of my music library is still made of MLP or Touhou fan music.

My other main hobby is puzzlehunts. I’ve solved them for around 10 years, and I’ve helped write multiple hunts, including MIT Mystery Hunt 2023. A list of puzzles I’ve written is here.

If you’re here for my blog, you’re probably here for my post about why deep RL doesn’t work, which is by far my most popular, but while you’re visiting, why don’t you read my post about the Neopets economy?