About
Hi! I’m Viraj, a Ph.D. student at Massachusetts Institute of Technology (MIT) advised by Prof. Brian C. Williams. At CSAIL’s Model-Based Embedded and Robotics Systems Group (MERS) and the Embodied Intelligence community, I design learning-enabled planners that help robots reason safely in the wild.
Research Focus
- Scalable multi-agent planning. Exploiting planning frameworks that decompose large multi-agent problems into tractable subproblems and re-compose them through conflict resolution, precedence reasoning, and hierarchical task structures.
- Risk-aware autonomy. Developing risk-aware algorithms that maintain global safety guarantees through dynamic risk budgeting, iterative risk reallocation, and skill chaining under uncertainty.
- Learning-guided planning. Integrating diffusion models, vision-language models(VLMs), and reinforcement learning with search-based methods to accelerate long-horizon task completion, improve generalization, and enable data-efficient closed-loop control.
- Representation learning for coordination. Learning structured latent spaces that encode notions of safety, cooperation, and temporal consistency, enabling interpolation-based reasoning and scalable multi-agent policy adaptation.
Background
- Graduate Research Assistant, MIT EECS
- M.Sc. in Robotics, Carnegie Mellon University (2021) under Prof. Stephen Smith.
- B.Tech. in Computer Science and Engineering, IIIT Delhi (2019) mentored by Prof. Tanmoy Chakraborty and Prof. Ponnurangam Kumaraguru.
