XIRRA
About
About XIRRA
XIRRA was founded in Shenzhen in July 2026 by former Huawei Cloud large-model CTO Li Yin, with Nanyang Technological University professor Zhang Hanwang joining as co-founder and chief scientist. The team is building a Large Physics Model that learns physical-state transitions and action policies together, aiming to support long-horizon decisions across different robots and operating scenes.
The model programme is paired with XR data and evaluation platforms. The data layer processes robot trajectories, human demonstrations and failure cases, while the evaluation layer runs reusable scenarios, records execution and returns failures to the next training cycle. Seed and angel financings announced in September 2026 support pretraining, physical-interaction data, hiring and scenario validation.
Articles
Team
- Li Yin / 李寅 - Founder and Chief Executive Officer - Founded XIRRA and leads its Large Physics Model, training infrastructure and industrial commercialisation programme.
- Zhang Hanwang / 张含望 - Co-Founder and Chief Scientist - Co-leads XIRRA's model architecture and research programme for physics-native world and policy learning.
Timeline
Events
- 2026-09-23 - Funding - XIRRA discloses seed and angel financings
Dunhong Asset Management led, with Huakong Fund, Sanhua Holding, Ginkgo Valley Capital and other investors participating. The capital is intended for physics-model pretraining, real-world interaction data, hiring and scenario validation. The company reported a roughly US$500 million valuation, without allocating proceeds or a valuation separately to each round.
Source
Context
Comparable profiles
- Physical Intelligence
comparable profile - Robot Foundation Model Developer
- RLWRLD
comparable profile - Robot Foundation Model Developer
- ACE Robotics (大晓机器人)
comparable profile - Robot Foundation Model Developer
- Saturn Dynamics
comparable profile - Robot Foundation Model Developer
- Tommoro Robotics
comparable profile - Robot Foundation Model Developer
Shared workflows
- Ropedia
workflow adjacency - Embodied World Model Training
- Stanhope AI
workflow adjacency - Embodied World Model Training
Context tags
- Embodied AI world models - Stack Layer - https://zhidx.com/p/596555.html - exact entity source reviewed for this bounded context.
- Embodied world model training - Workflow - https://zhidx.com/p/596555.html - exact entity source reviewed for this bounded context.
- Robot foundation model developer - Peer Group - https://zhidx.com/p/596555.html - exact entity source reviewed for this bounded context.