Bingyi Kang is an AI researcher focused on building intelligent agents capable of assisting humans in the physical world, with a research agenda spanning computer vision, reinforcement learning, and multimodal intelligence. His work is driven by two central challenges: enabling machines to perceive the world with human-like spatial understanding, and equipping them to make decisions grounded in that perception. Kang is best known as the creator and leader of the Depth Anything series—a family of models including DA, DAv2, PromptDA, Video DA, and DA3—which collectively aim to achieve human-level depth perception. The project has gained significant traction in the research and developer communities, amassing over 24,000 GitHub stars and establishing itself as a leading effort in spatial intelligence.

Prior to his current role as a Founding Scientist at AMI Labs, Kang led the Spatial Intelligence research team at ByteDance Seed, where he worked on advancing perception systems and multimodal reasoning at scale. His earlier experience spans some of the world’s leading AI research environments, including Sea AI Lab, Facebook AI Research (FAIR), and academic institutions such as UC Berkeley and the National University of Singapore. These roles allowed him to bridge foundational research with real-world applications, particularly in areas requiring tight integration between perception and action.

Kang completed his PhD in Artificial Intelligence at the National University of Singapore, with a visiting research period at UC Berkeley, and holds a bachelor’s degree from Zhejiang University. Across both academia and industry, he has consistently focused on pushing the boundaries of spatial perception and embodied intelligence. His work reflects a broader shift in AI toward systems that can not only interpret the world visually, but also reason, plan, and act within it—laying the groundwork for the next generation of intelligent agents.