Embodied Intelligence Observer

Zeno AI芝诺机器人

LatestAngel (Jan 2026, 赛意产业基金)

Undisclosed (to be confirmed)

Shanghai Zeno Future Technology Co., Ltd.

A CMU-lineage embodied AI startup focused on 'whole-body dexterous manipulation', advocating third-person video learning for autonomous robot evolution, building Full-stack Physical AI

Embodied ModelsChina

数据截至2026-08-19

FoundedMay 2025 (business registration) / Jun 2025 (formal operation)
HQShanghai (per original notes; a Zhihu report says the HQ is in Hangzhou with an R&D center in Sydney, to be confirmed)
FounderWilliam Zhi, Bai Huiyuan
Products0 items
Funding Rounds1 rounds
Events4 items

About

Zeno AI was founded in 2025 by William Zhi and Bai Huiyuan; the company is Shanghai Zeno Future Technology Co., Ltd., with its HQ recorded as Shanghai (other reports place it in Hangzhou with an R&D center in Sydney, to be confirmed). The company argues robots should not focus only on 'hands' but achieve whole-body coordination of torso, dual arms, hands and base, and takes third-person video as the core data source, letting robots learn and evolve autonomously by watching human operation videos — reducing dependence on teleoperated data collection — while exploring a Full-stack Physical AI route integrating 'embodied brain + cerebellum + body'. The core founders are William Zhi, a postdoctoral researcher at the CMU Robotics Institute and assistant professor at the University of Sydney, and Bai Huiyuan, former VP of Alibaba Group, with a team from CMU, Zhejiang University, the University of Sydney, UIUC and other top institutions at home and abroad. In January 2026 it completed an Angel round invested by 赛意产业基金; in August 2026, 3 founding-team papers were accepted to IROS 2026, covering 3D Gaussian global map fusion and sim-to-real transfer modeling of soft robots. No products have been officially released; the company is at the R&D and demo stage (to be confirmed).

Technical Approach

Advocates 'whole-body dexterous manipulation': uniformly coordinating base pose and upper-limb fine manipulation rather than focusing only on 'hands'. Takes third-person video as the core data source, letting robots learn and evolve autonomously from human operation videos to reduce dependence on teleoperated collection. Builds a Full-stack Physical AI stack from data collection and model training to body control, exploring an integration path of 'embodied brain + cerebellum + body'.

Commercialization(2)

  • Early stage; launched 2026 campus recruiting in April 2026 (for mechanical, automation and computer majors, with internships in Hangzhou and Sydney) to accelerate team expansion and product R&D
  • Commercialization progress and orders to be disclosed

Position & Strengths(3)

  • A CMU-lineage embodied AI startup listed alongside Unitree, Galbot and 星海图 among the most-watched companies (per original notes)
  • Hard-core academic pedigree of the core founders (CMU postdoc + University of Sydney assistant professor), with 3 papers accepted to IROS 2026 and frontier research capability
  • Differentiated route: focuses on 'whole-body dexterous manipulation' and 'third-person video learning', cutting into the intelligent-evolution path for complex long-horizon manipulation tasks

Links3

Zeno AI | Embodied Intelligence Observer