Maya Guru

I'm a predoctoral researcher with the PRIOR team at the Allen institute for AI (AI2) in Seattle, WA.

Previously, I was receiving my BSE in computer science at the University of Pennsylvania along with my MS in robotics at the GRASP Lab. After graduating, I joined the RAI Institute as a software engineer, working in various areas spanning embedded software, manipulation, and perception.

My research interests include scaling up robotic manipulation capabilities, with particular emphasis on whole-body manipulation and contact-rich tasks. My research aims to improve robot generalization, robustness, and adaptability by leveraging data-driven approaches that enable robots to perform effectively in new and unseen environments.

Email  /  CV  /  Twitter  /  LinkedIn  /  Github

News

Publications

MolmoBot is a robotics data-generation engine and policy suite that is trained exclusively on simulated expert trajectories generated in MolmoSpaces, outperforming state-of-the-art policies on tabletop pick-and-place and demonstrating sim2real transferability of mobile manipulation tasks.

MolmoSpaces is a fully open embodied AI platform featuring diverse indoor scenes, object models, and grasp annotations compatible with MuJoCo, Isaac Lab, and ManiSkill, along with benchmarks designed to enable distributional analysis of robot policies across systematic variations in environment and task conditions.

A motion planner for dexterous and whole-body manipulation that can be used to generate data across multiple embodiments, enabling efficient Sim2Real RL for complex manipulation tasks.

Website adapted from here.