BlockBeats News, July 27th, Physical AI data engine company Axis Robotics announced the completion of a $12 million seed round financing, led by Hack VC, with participation from Nomad Capital, Pi Network Ventures, 10K Ventures, and several angel investors.

Axis Robotics stated that this funding round will be used to build a large-scale, human-robot collaborative global robot data engine to address the data bottleneck in the Physical AI field, including challenges such as scarce training data, insufficient model generalization capabilities, and data fragmentation among different robot hardware.

Unlike large language models that rely on massive internet text data, Physical AI requires a large amount of real human-physical interaction trajectory data for training. Axis Robotics founder Chris stated that the company is building a sustainable data production system to accelerate the development of general AI by continuously generating, collecting, and optimizing robot training data.

Axis's "Composite Data Engine" integrates task generation, data collection, model training, and optimization processes, including:

Randomly generating data tasks with different objects, spatial layouts, visual environments, and robot morphologies through a task generation engine;

Providing a browser-based robot simulation remote operation platform to improve data collection efficiency;

Collecting real-world human action data through a mobile application;

Automating trajectory cleaning, domain randomization, and language labeling to generate multimodal datasets suitable for model training.

The company stated that it has already established a global robot data network composed of over 100,000 active contributors, capable of generating over 1,200 hours of simulation data and over 20,000 hours of real-world first-person view data per month.

Axis Robotics claimed that its dataset has demonstrated performance advantages in robot benchmark tests. In the LIBERO-Plus test, the π0.5 model trained on Axis's diverse dataset achieved a 4.9 percentage point increase in success rate compared to the RoboCasa365 benchmark dataset, which saw a 31.3 percentage point improvement.

In terms of commercialization, Axis Robotics has partnered with companies such as Booster Robotics, Manycore Tech, Feagine Robotics, Dexmal, Lotus Car, and Geely Auto to provide customized training data services to robot manufacturers, Physical AI model companies, and industrial automation enterprises.

The company stated that it will utilize the current funding round to enhance its data generation capabilities, expand the global contributor network, and further develop the core data infrastructure that supports next-generation artificial general intelligence.