Valts is a second-year PhD student in computer science, co-advised by Yoav Artzi in Cornell Tech. He applies machine learning techniques to robotics problems, attempting to scale them beyond instrumented environments and towards more complex real-world scenarios. Currently he trains robots to follow natural language instructions and perform manipulation behaviors in simulation, with the goal of transfer to a real robot. This involves using reinforcement learning, imitation learning, transfer learning, domain adaptation and other learning methods, but also designing new models and algorithms that are uniquely tailored for robotics tasks. In the past he has also worked on rescue robots, stereo vision and participated in various robotics competitions.