From language and vision to robot actions
A place to explore how robot models connect visual observations and language instructions to actions in the physical world.
Bio
I’m Kun Li, a robotics and AI scientist based in Greater Seattle. My interests span robot perception, robot learning, large language models, and building practical AI systems.
I’m currently Director of AI at Abby Care. Previously at Amazon, I worked on agentic AI and production machine learning systems. At Amazon Robotics, I designed and built 3D perception for the Robin robotic sorting system.
I earned my Ph.D. in Electronic Engineering at the Chinese University of Hong Kong and was a postdoctoral scholar at Caltech, working with Joel Burdick on robot learning and inverse reinforcement learning.
This site is a place for my writing, findings, and experiments.
kunliname@gmail.comRobot Foundation Models
Reading notes, questions, and experiments on learning general-purpose robot behavior.
Learning across tasks and embodiments
Questions about robot datasets, transfer between platforms, and how to evaluate generalization beyond familiar settings.
Sample previews for the layout. Original writing and reading notes will appear here.
3D Perception
Notes on geometry, spatial understanding, and perception for robotics.
Seeing the scene, understanding the task
A space for findings on 3D representations, pose estimation, sensor fusion, and the information a robot needs to act.
Sample preview. Original notes will appear here.
Agentic Systems
Writing and findings on agents, tool use, planning, and multi-step workflows.
From reasoning to useful workflows
A place for observations on tool use, model evaluation, and experiments with agents that carry out multi-step tasks.
Sample preview. Original notes will appear here.