RoboBoom
Machina Summit· July 22, 2026· 20:31

Keynote by Yuli Zhao (Galbot) on Building Generalist Robots for the Real World | MACHINA 2026

Yuli Zhao of Galbot presents at MACHINA 2026 on the company’s foundation models, data strategy, and deployments of generalist robots in manufacturing, retail, logistics, and health care.

In this talk

  1. 0:41

    Galbot’s foundation model and data

    Zhao describes Astro Brain as a model that brings together high-level reasoning, whole-body movement, and hand control. Galbot says Astro Data includes 20 billion pieces of embodied AI data, gathered through real-world operations, human behavior, teleoperation, and simulation. Zhao says combining synthetic and real-world data helps Galbot train robots for tasks that are difficult to capture directly.

  2. 5:47

    Robot skills for dynamic tasks

    Zhao presents a humanoid tennis demonstration that Galbot says runs autonomously, using a foundation model to track the ball, move into position, and return it. The training used motion-capture recordings of people playing tennis in a simulator. Zhao also describes dexterous-hand work, including manipulating varied objects and rotating two chestnuts as their balance shifted.

  3. 10:19

    Whole-body control and world action models

    Galbot’s architecture combines multimodal inputs with cerebrum and cerebellum components. Zhao says its whole-body control model was trained on two billion frames of human behavior and can perform movements beyond those shown in its training examples. Astro Brain World Action Model 0.5 supports longer, multitask activity, such as cooking, and Galbot says its 1.6-billion-parameter model can run on edge devices and across several robot platforms.

  4. 16:43

    Deployments and industrial partnerships

    Zhao names Bosch, Siemens, CATL, Mercedes-Benz, and Hyundai Motor among Galbot’s partners. He describes a FamilyMart robot that understood a spoken sausage order and delivered it, along with more than 100 smart pharmacies operating across 33 cities in China. Galbot also reports 170 humanoid-operated Gable stores in 40 cities and factory deployments, including Galbot S1 at Siemens and CATL.

Briefs summarise talks and interviews; timestamps open the video at that moment. Claims in a talk are the speaker's own.

More talks