Kaylee Burns

Google (United States), Stanford University

Papers

Total Citations

H-Index

About

Kaylee Burns is a leading researcher in robot learning, whose work is fundamentally reshaping how robots acquire and execute complex manipulation skills. Her primary research areas span end-to-end robot learning, action representation, and scalable data collection for imitation learning. Burns made a major contribution with her highly influential paper, "Implicit Kinematic Policies: Unifying Joint and Cartesian Action Spaces in End-to-End Robot Learning" (14 citations), which revealed that the choice of action space—whether joint positions or Cartesian end-effector poses—can dramatically impact robot performance, and proposed a unified framework to bridge this gap. She further advanced the field with "GenCHiP: Generating Robot Policy Code for High-Precision and Contact-Rich Manipulation Tasks" (5 citations), demonstrating how Large Language Models can be harnessed to generate code for tasks requiring precise force reasoning, a previously unsolved challenge. Most recently, her work "RoboCrowd: Scaling Robot Data Collection Through Crowdsourcing" (2025) tackles the critical bottleneck of data scarcity by enabling large-scale, non-expert data collection, promising to democratize robot training. Through these contributions, Burns is driving a paradigm shift toward more capable, data-efficient, and accessible robotic systems.