Shivam Aarya

Hey there! I'm Shivam, a Ph.D. student in the RIPL Lab with Zsolt Kira.

My Ph.D. research focuses on fully autonomous robot learning - specifically, overcoming the laborious process of teaching robots new tasks, which remains one of the major bottlenecks holding robotics back. The diversity of robot embodiments, tasks, configurations, and environments makes this learning process especially difficult to scale. I aim to make robot learning fully autonomous by eliminating the need for human demonstrations, minimizing the sim-to-real gap, and automating reset policies. Ultimately, I want to shift the focus from minimizing human effort to optimizing the cost and latency of robot learning.

I began my research as an undergraduate at Johns Hopkins University with Prof. Tianmin Shu's SCAI Lab in collaboration with Prof. Josh Tenenbaum at MIT. My research focused on enabling intuitive cooperation between people and robots so that robots can help everyone live better lives. In GOMA, we developed methods for general-purpose robots to correctly interpret human commands and reliably complete their tasks. In SIFToM, we explored how robots can build comprehensive world models that are tightly synchronized with the mental models of the humans they assist, understand the physical environments they inhabit, and deeply model the objects they manipulate. I also joined the Computational Cognitive Science (CoCoSci) Lab at MIT as a Research Assistant to develop CogGym, a platform to standardize and host hundreds of online human-agent interaction studies.

Beyond research, I really enjoy building interactive and intuitive systems, which often pulls me into the startup world. I was involved with three student-led startups, two of which were funded by JHU's Spark Accelerator, and the third was supported by the Johns Hopkins Office of Undergraduate Research (HOUR).

If you're into AI, robotics, or just want to chat about cool projects, feel free to reach out!