John (Jack) Friess
/
i recently graduated from ut austin where i studied cs + math in the turing scholars honors program. i like working at the center of product, infra, and research. i am currently working through jacob hilton's deep learning curriculum, you can see my progress.
a few highlights about me:
inspired by rocket league, i built a playable 3D car soccer game in C++ with my colleague beto as our final project for our graduate physical simulation class. we developed a custom physics engine with rigid body simulation, collision detection, and collision response. this project also won us the Lapin d'Or, which is awarded by Dr. Etienne Vouga for the best project in the class.
play using WASD to drive, Space to jump, Left Mouse to boost, and Right Mouse to control the camera. press ESC to exit.
Towards Robust Policy Extraction from Universal Value Functions — advised by Dr. Amy Zhang. universal value functions (uvfs) are a way to teach robots broad notions of progress from unlabeled video, then transfer that knowledge to new tasks. my thesis proposes a new method for extracting a strong policy from a pretrained uvf when only limited task data is available.