Hersh Sanghvi

Hello! I’m a 5th year Computer Science Ph.D student in the GRASP Lab at the University of Pennsylvania, advised by Dr. Camillo Jose Taylor. My research interests center on using learning to adapt control and planning algorithms for robots to new situations. Current and past projects include developing approaches for fast controller adaptation to new robots and environments, supervised learning for footstep planning for legged robots, and using model-free reinforcement learning for legged robot navigation. If you’re curious about my work, you can check out my publications and writing, and other projects I’ve worked on.

Besides robotics, I’m also interested tech policy and am treasurer of the Penn Science Policy and Diplomacy Group. As a member of PSPDG, I’ve helped develop workshops on writing science policy memos, presenting science policy ideas to legislators, and recognizing gerrymandering. I’ve also published policy memos and blog articles on disinformation on social media, regulation of AI, and secure voting machines (check them out in the “Publications and Writing” tab).

I obtained my BS in Electrical and Computer Engineering from the University of California, Berkeley in 2019. I got my start in research working with Dr. Roy Ben-Shalom on NeuroGPU, a method for speeding up neuronal simulations by using GPU acceleration. Starting in my sophomore year, I worked in the Biomimetic Millisystems Lab lab with Dr. Justin Yim on developing control and state estimation algorithms for Salto, a one-legged hopping robot. I was also on Berkeley’s solar vehicle team, CalSol, for which I designed PCBs and wrote firmware for vehicle electronics, and implemented filtering algorithms for battery state estimation.