MiraiBit
About
About MiraiBit
MiraiBit develops data and model-development infrastructure for embodied AI. Its OpenARM-based workflow combines teleoperation, synchronized robot-data recording, playback and quality control, with export into LeRobot-compatible formats for policy training and evaluation. OpenARM is used as an internal and customer-facing data-collection environment rather than presented as a MiraiBit robot product.
The company also works on VLA models, simulation-to-real transfer and validation on physical hardware. In September 2026 it disclosed active manipulation-data collection on OpenARM 2.0 and the addition of Unitree G1 to its humanoid development environment. MiraiBit also provides autonomous-driving and broader AI engineering services.
Observed use
- Anvil OpenARM 2.0 at MiraiBit Physical-AI R&D environment
2026-09-07 - Research Experiment - Trained With - MiraiBit operates OpenARM 2.0 as a physical teleoperation and manipulation-data collection environment for embodied-AI training and evaluation.
Source - G1 at MiraiBit Physical-AI R&D environment
2026-09-07 - Research Experiment - Tested With - MiraiBit added a physical Unitree G1 to its development environment for humanoid Physical-AI experiments.
Source
Team
- Shi Gengtian - Representative Director - Current representative director of Mirai Bit Co., Ltd.
Timeline
Events
- 2026-09-07 - Capability - MiraiBit starts real-robot Physical AI development
An OpenARM 2.0 teleoperation environment is collecting robot-operation data for AI training and evaluation. The workflow continues from data collection through model development to validation on physical hardware. Unitree G1 has been added to the development environment for humanoid experiments.
Source
Context
Comparable profiles
- Physical Intelligence
comparable profile - Vla Robot Policy Developer
- X Square Robot
comparable profile - Vla Robot Policy Developer
- PHYMI
comparable profile - Vla Robot Policy Developer
Shared workflows
- Lightwheel
workflow adjacency - Robot Learning Data Collection
- Synthium
workflow adjacency - Robot Learning Data Collection
- Mind Robotics
workflow adjacency - Robot Learning Data Collection
- SenseGlove
workflow adjacency - Robot Learning Data Collection
- MANUS
workflow adjacency - Robot Learning Data Collection
- Trossen Robotics
workflow adjacency - Robot Learning Data Collection
- Rokoko
workflow adjacency - Robot Learning Data Collection
- StretchSense
workflow adjacency - Robot Learning Data Collection
- Maniformer
workflow adjacency - Robot Learning Data Collection
- Jiiov / 极豪科技
workflow adjacency - Robot Learning Data Collection
- Hub
workflow adjacency - Robot Learning Data Collection
- Physen AI
workflow adjacency - Robot Learning Data Collection
Context tags
- VLA robot policy developer - Peer Group - https://robotstart.info/article/2026/09/07/382378.html — exact reviewed source supports this bounded company context.
- Physical AI training data - Stack Layer - https://robotstart.info/article/2026/09/07/382378.html — exact reviewed source supports this bounded company context.
- Robot learning data collection - Workflow - https://robotstart.info/article/2026/09/07/382378.html — exact reviewed source supports this bounded company context.
- Robot-learning infrastructure stack - Stack Layer - https://www.mirai-bit.co.jp/ — exact reviewed source supports this bounded company context.