Matt Deitke

Matt Deitke

Hi! I'm a researcher at the Allen Institute for AI and an undergraduate at the University of Washington in the Paul G. Allen School of Computer Science. I am interested in general solutions to problems in computer vision, reinforcement learning, and embodied AI.

Previously, I worked as a computer graphics developer in the Department of Athletics at The Ohio State University and the University of Cincinnati, along with several other startups. While in high school, I studied machine learning at Georgia Tech.



A benchmarked visual navigation environment within AI2-THOR consisting of scenes in both simulation and real life.

The primary objective is to train agents in a simulation environment that can then act effectively in the real world.

Learn More


An open source, physics simulated, 3D library facilitating simulation to reality transfer research in embodied AI across many different tasks.
Install the Library


CVPR 2020 Embodied AI

The RoboTHOR Challenge is being organized at CVPR 2020 as part of the Embodied AI Workshop.

Participants submit trained visual navigation models, which will be evaluated using a real robot in physical apartments

Get Started

Read about it in


CVPR Pretty Papers [code]: Favorites: Course Notes: