Physical AI data engine company Axis Robotics said on July 27 that it has raised $12 million in a seed round led by Hack VC. Nomad Capital, Pi Network Ventures, 10K Ventures, and multiple angel investors also joined the round.
The company said the capital will go toward building a large-scale, human-in-the-loop global robot data engine to address data bottlenecks in Physical AI, including a shortage of training data, limited model generalization, and fragmented datasets across different robot hardware systems.
Funding targets core data constraints in Physical AI
Axis Robotics said Physical AI has a different data requirement from large language models, which are trained on massive volumes of internet text. For robotics systems, the company said, training depends on large amounts of real human physical interaction trajectory data.
Founder Chris said Axis is building a sustainable data production system that continuously generates, collects, and optimizes robot training data to speed up the development of general-purpose robot intelligence.
Composite data engine spans generation, collection, training, and optimization
Axis said its "composite data engine" combines task generation, data collection, model training, and optimization in one workflow. The system includes:
- a task generation engine that randomly creates data tasks across different objects, spatial layouts, visual environments, and robot forms;
- a browser-based remote teleoperation simulation platform designed to improve data collection efficiency;
- a mobile application used to capture real-world human motion data;
- an automated pipeline for trajectory cleaning, domain randomization, and language annotation to produce multimodal datasets for model training.
More than 100,000 active contributors in its global network
The company said it has already built a global robot data network with more than 100,000 active contributors. That network generates more than 1,200 hours of simulation data each month and over 20,000 hours of real-world first-person data.
Axis Robotics said its datasets have shown a performance edge in robot benchmarking. In LIBERO-Plus, the π0.5 model trained on Axis's diversified dataset improved its success rate by 4.9 percentage points, and by 31.3 percentage points compared with the RoboCasa365 benchmark dataset.
Commercial partnerships already in place
On the commercial side, Axis Robotics said it has partnered with Booster Robotics, Manycore Tech, Feagine Robotics, Dexmal, Lotus Car, and Geely, among others. The company provides customized training data services for robot manufacturers, Physical AI model companies, and industrial automation firms.
Axis said it plans to use the new funding to expand its data generation capacity, grow its global contributor network, and keep building the core data infrastructure it says will support the next generation of general-purpose robot intelligence.