Tony Zhao

Zhao studied electrical engineering and computer sciences at the University of California, BerkeleyUniversity of California, BerkeleyThe University of California, Berkeley (UC Berkeley) is a campus of the University of California system; the University of California was founded in 1868 through a merger of two institutions and moved to Berkeley, California, in 1873.Open the full entry, advised by Sergey LevineSergey LevineSergey Levine is a computer scientist and associate professor of electrical engineering and computer sciences at UC Berkeley who works on deep reinforcement learning and robot learning, co-developed algorithms such as Soft Actor-Critic and Conservative Q-Learning, and in 2024 co-founded the robot foundation model company Physical Intelligence.Open the full entry and Dan Klein, and received his bachelor’s degree in 2021.⁠[1][2]

In October 2020 he released code for MELD, a method for meta-reinforcement learning from images via latent state models, on which he was co-first author with Anusha Nagabandi. In February 2021 he released code for contextual calibration; the accompanying paper, “Calibrate Before Use: Improving Few-Shot Performance of Language Models,” co-first-authored with Eric Wallace and published at ICML 2021, introduced a data-free procedure that improves GPT-2 and GPT-3 few-shot accuracy by up to 30 percentage points.⁠[1]

Zhao pursued a Ph.D. in computer science at Stanford University, advised by Chelsea FinnChelsea FinnChelsea Finn is a computer scientist and robot-learning researcher who is an assistant professor of computer science and electrical engineering at Stanford University and a co-founder of the robotics AI company Physical Intelligence. Known for work on meta-learning and robot learning, she won the 2018 Doctoral Dissertation Award of the Association for Computing Machinery and was named to the TIME100 AI list in 2026.Open the full entry. In September 2022 he received the Stanford Robotics Fellowship for 2022–23.⁠[1]

In 2022 his paper with Siddharth Karamcheti, Thomas Kollar, Chelsea Finn and Percy Liang, “What Makes Representation Learning from Videos Hard for Control?”, was a Best Paper Award finalist at the RSS 2022 Workshop on Scaling Robot Learning.⁠[1]

On March 27, 2023, Zhao released the code for ALOHA and ACT. On April 23 his paper with Vikash Kumar, Sergey Levine and Chelsea Finn, “Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardware,” was submitted to arXiv; it was later published at RSS 2023.⁠[1][3]

ALOHA (A Low-cost Open-source HArdware) is a low-cost open-source hardware system for bimanual teleoperation, and ACT (Action Chunking with Transformers) is an imitation-learning algorithm that learns a generative model over action sequences. The paper reports that with about 10 minutes’ worth of demonstrations, ACT allowed a robot to learn six difficult real-world tasks, such as opening a translucent condiment cup and slotting a battery, with 80–90% success.⁠[3][1]

On January 3, 2024, Zhao and Zipeng Fu, as project co-leads, released the hardware and learning code for Mobile ALOHA, with Chelsea Finn as advisor; the paper was later published at CoRL 2024. Mobile ALOHA augments ALOHA with a mobile base and a whole-body teleoperation interface; the paper reports that with 50 demonstrations per task, co-training with static ALOHA data increases success rates by up to 90%, allowing the robot to autonomously complete mobile manipulation tasks such as sautéing and serving a piece of shrimp.⁠[4][1]

On September 12, 2024, Google DeepMindGoogle DeepMindGoogle DeepMind is Google’s artificial intelligence research organization, formed in April 2023 by merging DeepMind, which Google acquired in 2014, with the Brain team from Google Research, and led by CEO Demis Hassabis. It developed AlphaGo, AlphaFold, and the Gemini family of models, and Hassabis and John Jumper received the 2024 Nobel Prize in Chemistry for AlphaFold’s protein structure prediction work.Open the full entry presented ALOHA Unleashed, authored by Zhao, Jonathan Tompson, Danny Driess, Pete Florence, Kamyar Ghasemipour, Chelsea Finn and Ayzaan Wahid. The method builds on the ALOHA 2 platform, which was based on the original ALOHA from Stanford University; with it, a robot learned to tie a shoelace, hang a shirt, repair another robot, insert a gear and clean a kitchen. The paper was published at CoRL 2024.⁠[5][1]

In 2024 Zhao dropped out of his Stanford computer science Ph.D. program. Sunday says his decision to leave academia was driven by the conviction that robotics research had matured enough to be translated into real-world impact.⁠[2][1]

Zhao co-founded the home-robotics company Sunday Robotics with Cheng Chi, serving as chief executive while Chi is chief technology officer. The two began working in a garage, using 3D printers to develop the robot’s hardware and appearance. On November 19, 2025, the company publicly unveiled its household robot Memo.⁠[6][7]

At the launch Zhao said: “We built Memo to give people back time for what matters, with the safety needs for any family in mind. This is a turning point for home robotics.” He also said: “We want to build robots that free people from laundry, from the dishes, from all chores.”⁠[6][7]

On training data, Zhao said: “If you think about the most powerful AIs, ChatGPT or image-generation models, they are trained on the whole internet. We just don't have the internet for robotics.” He said the gloves Sunday uses to collect data cost roughly $400 a pair and provide a more accurate training signal than teleoperation, and that “people should be able to teach their own robots.”⁠[7]

On March 12, 2026, Sunday announced a $165 million Series B at a $1.15 billion valuation, led by CoatueCoatueCoatue is a technology- and innovation-focused investment management firm founded by Philippe Laffont in 1999, operating in both public and private markets. Its current website lists Thomas Laffont as co-founder and chief investment officer of private investments; a Form ADV filed on July 29, 2026, disclosed regulatory assets under management of $92,705,722,582.Open the full entry, bringing its total funding to $200 million. Zhao said: “We raised our Series B to stop giving demos. Now, we’re focusing entirely on deployment, with beta deliveries starting in just months.”⁠[8]

Zhao interned at Tesla Autopilot and Google X Intrinsic and worked as a part-time student researcher at Google DeepMind.⁠[1][2]

Sunday says Zhao leads a team of more than 140 engineers, designers and researchers in Silicon Valley and that he credits the team’s breakthroughs to one core value: persistence.⁠[2]

Personal website: https://tonyzhaozh.github.io/ (opens in a new window).⁠[1]

X: Public account: https://x.com/tonyzzhao (opens in a new window).⁠[1][2]

LinkedInLinkedInLinkedIn is a social networking platform for professionals. It was founded in 2002 by Reid Hoffman and others, launched in 2003, and listed on the New York Stock Exchange in 2011; after Microsoft acquired it for approximately $26.2 billion in 2016, it became a wholly owned Microsoft subsidiary. Its revenue comes from membership subscriptions, advertising, and recruitment solutions.Open the full entry: Public profile: https://www.linkedin.com/in/tony-z-zhao/ (opens in a new window).⁠[1]