Mohamed Elsayed

I am a PhD student at the RLAI lab, working with Rupam Mahmood. My research focuses on learning from a stream of experience. Humans and other animals keep improving through their interaction with the world, using each moment as it happens rather than recording it to revisit later. My goal is to build artificial agents that can do the same.

Today's deep learning systems, in contrast, depend on stored experience to retain knowledge, stabilize updates, and assign credit over time. I develop simple methods that do all three from the stream itself, spanning (1) continual learning without forgetting or losing the ability to learn, (2) deep reinforcement learning that learns reliably from one transition at a time, and (3) recurrent memory that learns from delayed feedback without backpropagating through the past. These methods are efficient enough for learning on real robots and even microcontrollers.

Prior to this, I completed my master's degree from the University of Alberta and my bachelor's degree in Communication and Information Engineering from Zewail City of Science and Technology, Egypt.

Outside of research: I'm passionate about photography. I also enjoy observing animal behavior and intelligence, particularly ravens.