Summary
Cheng Chi is the co-founder and Chief Technology Officer of Sunday Robotics, a Mountain View startup building Memo, an autonomous home robot for household chores. Before co-founding Sunday, Chi was a robotics PhD researcher at Columbia University and Stanford University under advisor Shuran Song, where he authored Diffusion Policy and the Universal Manipulation Interface (UMI)—two influential contributions to how robots learn manipulation skills from human demonstration data.
Career
Chi earned a bachelor’s degree in computer science from the University of Michigan. He began his robotics career in industry, joining autonomous vehicle company Nuro’s mapping and localization team in January 2020, where he worked on classical robotics problems before shifting toward machine learning.
In January 2021, Chi began his PhD in computer science at Columbia University, joining the Columbia Artificial Intelligence and Robotics Lab under Professor Shuran Song, maintaining a 4.02 GPA. His early research explored geometric and probabilistic algorithms, deformable object tracking, and garment manipulation, including work on GarmentNets (category-level pose estimation for garments) and DextAIrity (deformable manipulation), both presented at Robotics: Science and Systems (RSS) and earning Best Paper Award finalist recognition. He also completed research internships at Toyota Research Institute in 2022 and 2023, studying imitation learning for contact-rich manipulation tasks.
In early 2022, as diffusion models began transforming image generation with tools like DALL-E 2 and Stable Diffusion, Chi recognized an opportunity to apply the same technique to robotics. He authored Diffusion Policy: Visuomotor Policy Learning via Action Diffusion, presented at RSS 2023, which showed that denoising diffusion models could generate more robust robot action trajectories than prior approaches, particularly for tasks with multimodal or high-dimensional action spaces. Chi has described this as “one of the first papers that brought the diffusion model into robotics” and the technical foundation for his subsequent work.
When Professor Song moved to Stanford University to join the faculty full-time in 2023, Chi moved with her, continuing his PhD research as a Student of New Faculty in Stanford’s Robotics and Embodied Artificial Intelligence Lab. During a research internship at Toyota Research Institute’s Large Behavior Model team in mid-2023, he developed the Universal Manipulation Interface (UMI), a low-cost, hand-held gripper system designed for scalable, in-the-wild data collection that mirrors a robot’s own end-effector, allowing human demonstrations to translate directly into robot training data. Chi has explained that the UMI project represented a paradigm shift in his approach to robotics: rather than engineering separate perception, planning, and control systems, a single neural network trained on UMI-collected data could learn to perform all three functions together.
In 2024, Chi connected with Tony Zhao after the two noticed each other’s research publications had appeared around the same time. They began prototyping together informally—clamping a robot arm to a desk in Chi’s apartment—before formally co-founding Sunday Robotics in May 2024, with Chi serving as CTO. At Sunday, Chi has applied the data-collection philosophy behind UMI to the company’s proprietary Skill Capture Glove, which similarly mirrors the robot Memo’s own hand and sensor configuration to collect transferable manipulation data from a distributed network of human demonstrators, without relying on in-home teleoperation.
As CTO, Chi has led Sunday’s technical development through the company’s stealth period, its public launch of Memo in November 2025, and its rise to a $1.15 billion valuation following a $165 million Series B in March 2026. He has discussed the company’s engineering philosophy and data pipeline publicly, including in an interview with The Robot Report and an appearance on the No Priors podcast alongside Zhao.
Source Notes
This profile draws on Cheng Chi’s personal academic website and CV, his LinkedIn profile, and an interview with The Robot Report discussing his UMI gripper and Diffusion Policy research. His publication record—including Diffusion Policy, UMI, GarmentNets, and DextAIrity—is corroborated by his personal website and academic conference proceedings (RSS, ICCV, IROS).