Matthew seeks to rethink the fundamental components of today's learning agents with hopes of creating the algorithmic machinery needed for agents to tackle more challenging, opening-ended domains. His research involves feature induction for agents with low level sensory input, model based reinforcement learning, and unsupervised exploration. Matthew also moonlights as a traffic engineer (highwayman?), investigating novel extensions of the Autonomous Intersections Management (AIM) project. In his spare time he enjoys guitar, weightlifting, foosball, bridge, and yoga.