Embodied AI Data Startup Secures Funding from Robot Giants

A burgeoning Chinese venture, ZhiYu Jishi, is poised to revolutionize data infrastructure for embodied artificial intelligence, having successfully closed an early-stage funding round. The capital infusion, described as tens of millions of yuan, was strategically backed by a consortium of four prominent robotics manufacturers: LingChu Intelligence, QiongChe Intelligence, Zhejiang Humanoid, and X Square. This significant backing from industry leaders underscores the critical need for advanced data solutions in the rapidly evolving robotics landscape.

The company's mission is to serve as the essential data gateway and training input provider for the embodied AI sector. ZhiYu Jishi aims to transform vast quantities of raw, unstructured data from the physical world into meticulously curated training datasets. These datasets are specifically optimized to enhance robot task success rates, addressing a key bottleneck in current AI development for physical systems. The market for robotics in China has experienced a surge, with investments exceeding 10 billion yuan in the current year alone, highlighting the opportune timing for ZhiYu Jishi's innovative approach.

Leading the charge is CEO Yang Zhexuan, a former key architect at PingCAP, bringing deep expertise in large-scale distributed systems. He is complemented by CTO Xu Liangwei, whose background spans Tencent and XPeng Robotics, offering invaluable practical knowledge in robot hardware, software, and embodied intelligence algorithms. The operational strategy is guided by COO Zhang Jiye, a former Huawei executive and ecosystem lead at investor QiongChe Intelligence, ensuring a robust go-to-market and partnership framework.

ZhiYu Jishi has ambitious plans to establish extensive data collection facilities across China, spanning over 10,000 square meters and housing more than 400 robots of diverse form factors. The company projects accumulating an impressive 200 petabytes of heterogeneous data by the close of 2026. This massive data repository will fuel the development of more sophisticated and capable AI-driven robots.

Central to their offering is an automated data compilation pipeline designed to overcome the inherent challenges in robotics training data. This includes a proprietary quality inspection system leveraging distributed computing, which promises to significantly reduce the noise often present in traditional training sets. Furthermore, the company is developing a unique heterogeneous data lakehouse architecture tailored for embodied intelligence and a semantic search engine for efficient data retrieval, enabling developers to pinpoint specific scenarios and skills for targeted training.

Looking ahead, ZhiYu Jishi intends to expand its data collection capabilities by incorporating first-person perspective (Ego-Centric) data, aligning with methodologies like NVIDIA EgoScale. This expansion, slated for Q2 2026, will capture crucial pre- and post-training data inputs for embodied AI models. The company's early commercial traction is already evident, with existing orders approaching 100 million yuan, signaling strong market validation for their data-centric infrastructure solutions.