Research Interests
I am broadly interested in learning and control in both robots and brains, as well as robotics for brains (such as brain-machine interfaces and prosthetics) and brains for robots (biologically-inspired algorithms and architectures). My work largely draws on tools in deep learning, optimal control, reinforcement learning, graphical models, and computational neuroscience. Some overarching research goals of mine are to:

Understand what the right amount of prior structure and domain knowledge is to incorporate in learning and control algorithms and how to optimally leverage it to improve the speed of learning and generalization

Investigate how population-level neural computations perform decision making and generate movements, as well as how neurons co-adapt their behavior during learning and the underlying rules which guide this process

Translate these findings to solve real-world problems in robotics, neuroscience, and medicine

On the robotics side, my focus has been on integrating learned components within the framework of model predictive control (MPC) to improve controller performance and reduce computational requirements. In the neuroscience domain, I am applying recurrent neural networks as a test-bed to investigate these questions and analyzing recordings of neural populations using deep generative models.