All Recordings

Structure or Scale: How Much Should We Engineer into Robot Learning?

About This Talk

Learn the tradeoffs between end-to-end robot learning and modular pipelines. A panel of engineers discusses how much structure to engineer into robot learning systems, scaling robot AI, and whether learned or engineered approaches work best. Featured at Actuate, Foxglove's annual conference for robotics developers.

Speakers

Evan Morikawa, Member of Technical Staff, Generalist

Evan led Applied Engineering at OpenAI and now does whatever is necessary at Generalist to realize general purpose robots. When he's not thinking about AI or robots, he chases around 2 small children and is starting to collect radioactive rocks.

Shitij Kumar, Technical Manager, Motion Planning and Simulation, Dexterity

As a roboticist and engineering leader, Shitij Kumar guides the development of core motion and grasping intelligence for AI-powered robots at Dexterity, Inc., a company developing AI robots with human-like dexterity for logistics applications. He earned his Ph.D. in Engineering from the Rochester Institute of Technology, where his research focused on AI applications in industrial human-robot collaboration. During his doctoral work, he partnered directly with industry to design and deploy production-level robotic solutions. He brings expertise in AI, controls, simulation, motion planning, and robotic solution design to advance Physical AI in automation, with a focus on creating robotic systems that address labor challenges and enhance human capabilities in physically demanding environments.

John Betancourt, Head of AI and Robotics, Robot.com

Head of AI and Robotics at Kiwibot, driving innovation in autonomous systems and mobile robotics at one of the fastest-growing last-mile delivery companies worldwide.

More Recordings