Todor Markov

About me

I am a Machine Learning Fellow at OpenAI, working on the Multiagent team. Before that, I was a software engineer at Blend.

I graduated from Stanford in June 2017 with a bachelors in Symbolic Systems and a masters in Statistics. My coursework was mainly focused on the the intersection of statistics and artificial intelligence, but I also studied a bit of philosophy and neuroscience. During my senior year, I did research in the Stanford AI Lab under Stefano Ermon on using machine learning to improve batteries.

My main research interest is in deep reinforcement learning (though I am starting to explore generative models & NLP a little bit). Some specific questions I want to figure out are: How do we create agents that can act competently in complex environments with long time horizons and sparse rewards? How do we create agents that can effectively generalize from experience and have those generalizations be transferable across environments or domains? Can we use reinforcement learning to get a better formal understanding of the principal-agent problem and potential mitigations for it?

In my free time, I like to practice brazilian jiu-jitsu, read, spend time in the effective altruism community, and generally try to figure out how the world works.