About me!

Hi there! ( ◕ ◡ ◕)

My name is Mahan Fathi (ماهان فتحی), I am a MSc student at Mila, advised by Pierre-Luc Bacon. I am also a student researcher at Google DeepMind Montreal, supervised by Ross Goroshin. It's hard for me to put a finger on what exactly my research interests are; pretty much all things ML interest me, so long as the research is done right. I am a strong refuser of the wrong "publish or perish" mindset, and have no interest in publishing uninspiring work. To me, the upshot of Reinforcement Learning research from the past few years is that RL needs a major overhaul. To this end, during my PhD, I aim to conduct my research on training agents that learn to solve MDPs by directly optimizing for the metrics that we we consider important for a typical RL algorithm, i.e. sample efficiency, safe exploration, etc. I believe recent advances in meta-learning and sequence modeling could potentially help us take the next leap in RL.

I received my bachelor in Mechanical Engineering from Sharif University of Technology in 2017. Towards the end of my studies I got involved in Robotics and Control, and found Deep Reinforcement Learning as a very powerful tool to tackle previously unsolved tasks in locomotion, etc. I then worked for 3 years as a Software Engineer, Data Scientist, and Team Leader in top tech companies of Iran, namely CafeBazaar (user-base of 50 millions) and TAPSI. I also spent a summer as a research assistant at Aalto University under the supervision of Perttu Hämäläinen, looking more into model-based RL.

For a more detailed summary of my experience please refer to my Curriculum Vitae.