Data Infrastructure for Embodied AI
On June 17, 2026, Mifeng Technology, an embodied AI data platform company incubated by Zhiyuan Robot, announced the completion of a hundreds-of-millions-yuan Angel+ strategic funding round.
The Data Bottleneck
Unlike large language models that rely on internet text and images, embodied AI model training depends on real physical world interaction data including robot grasping, assembly, force control, and visual perception. This data is significantly more expensive to collect.
MEgo Series: No-Robot-Body Hardware
Mifeng Tech's core offering is the MEgo series:
- MEgo View: Head-mounted and wrist-mounted dual-perspective device
- MEgo Gripper: Two-finger gripper recording action trajectories
Using infrared active light and VSLAM fusion positioning, the system achieves millimeter-level trajectory reconstruction with time-synchronized multimodal data alignment.
Scale Ambitions
The company plans to achieve 10 million hours of data collection capacity in 2026 and targets 10 billion hours by 2030, positioning itself as a critical infrastructure provider for the embodied AI ecosystem.
Background: Founded in February 2026, Mifeng Tech is a one-stop Physical AI data service platform. CEO Yao Maoqing is also a partner at Zhiyuan Robot.
