DexForce

A Shenzhen embodied-AI company that trains robots mainly on synthetic data from generative simulation.

DexForce was founded in Shenzhen in June 2021. Its founder, Kui Jia, is a professor at the Chinese University of Hong Kong, Shenzhen, with a long research background in deep learning and 3D geometric vision. Where most companies collect data by having a person teleoperate a real robot, DexForce takes a “generative simulation” approach: its own DexVerse embodied-AI engine generates virtual scenes and synthetic data in bulk to train models. Jia's reasoning is that physical variation — object positions, materials, environments — can all be generated in simulation, while “what to do,” semantic knowledge, still depends more on real data. Its other products include the DexWorldModel world model, the DexForce W1 Pro humanoid robot, and the DexSense vision-only sensor; the company says it is already deployed across more than 50 industry verticals. On June 30, 2026, it announced a Series B round of more than RMB 1 billion at a post-money valuation above RMB 10 billion, and is reportedly preparing an IPO.

ExampleJia has estimated that a single teleoperator can only collect 100-150 demonstrations a day, which is why DexForce chose to generate large volumes of synthetic data in simulation to cover physical-level generalization instead.

Also called

DexForce Technology

Related

Generative Simulation · Synthetic Data · Simulation Data · Sim-to-Real Transfer · World Model · Real-Robot-Data Camp vs. Sim-Data Camp

Sources

跨维智能官网：公司介绍 (Chinese)
跨维智能完成10亿元B轮融资 - DexForce (Chinese)
证券时报：跨维智能完成10亿元B轮融资将迎IPO (Chinese)

As of

2026-07