Tsinghua-affiliated Embodied Brain company completes tens of millions of yuan in financing and receives support from multiple industrial players | Exclusive report by Hardcore Kr

Author | Huang Nan

Editor | Yuan Silai

Yingke has learned that Beijing Qianjue Technology Co., Ltd. (hereinafter referred to as "Qianjue Technology") recently completed a Pre-A round of financing worth tens of millions of yuan. This round was jointly led by Zhuichuang Venture Capital and DT Capital, with a strategic investment from Jingye Intelligent Industry Fund. Maple Pledge Capital has been providing long - term financing advisory services. The funds will be mainly used for team expansion and global commercialization.

Founded in June 2023, "Qianjue Technology" benchmarks against the US startup Physical Intelligence. It focuses on the "robot brain" of embodied intelligence and specializes in the research and application of large decision - making and planning models. It enables robots of various forms to independently observe the environment, think, and act like humans. It helps robots break through the limitations of only performing "single tasks" in "specific scenarios" and enter markets such as household and commercial services.

Dr. Gao Haichuan, the founder and CEO, graduated from the Department of Automation at Tsinghua University. Since 2018, he has served as the team leader in the bionic dual - arm robot research group led by Professor Chen Feng, one of the first experts in brain - like computing research in China. He led the team to complete the design of several dual - arm autonomous decision - making robots from scratch.

By October 2024, "Qianjue Technology" had accumulated a pre - trained dataset of embodied perception and decision - making at the scale of hundreds of millions, covering a wide range of application scenarios. The company has published articles in top - tier global conferences and journals multiple times and, based on this, was the first to launch the world's first full - stack brain - like multimodal large model.

In current embodied intelligence research, the technological research on the "brain" and "cerebellum" has attracted much attention. That is, to make robots imitate the division of labor in the human brain's operation. Through the collaborative mechanism of the brain and cerebellum and the complementarity of the two structures, the brain is responsible for high - level perception and decision - making functions such as vision, hearing, and consciousness, while the cerebellum is responsible for coordinating data to control movement, balance, and behavioral postures. Building a general embodied robot brain is a key factor in promoting the entry of robots into a wider range of application scenarios.

With a brain - like technology route as its characteristic, "Qianjue Technology" has developed a large - scale product - level robot perception and decision - making model that can work fully autonomously and dynamically respond to environmental changes. For example, after integrating "Qianjue Technology"'s embodied brain, a household service robot can identify and judge objects and tasks in the environment. Users only need to issue a demand such as "tidy up the room," and the robot can enter the task - planning state on its own and complete tasks such as folding clothes, storing toys, and cleaning up garbage one by one, achieving true task generalization across forms, environments, tasks, and objects.

On the hardware side, "Qianjue Technology"'s brain - like multimodal large model can be adapted to various robot hardware, including humanoid and non - humanoid robots and other automated equipment. It operates in a decoupled manner from the perception, decision - making to the control links, with higher flexibility and greatly expands the application boundaries.

In terms of data, currently, "Qianjue Technology" uses the world's largest household scenario dataset to train the model. Compared with the traditional large - model technology route, it not only has a lower threshold for training data, which can significantly reduce the model training cost, but also can complete the generalization and migration of various physical environment variables and action patterns. At the same time, the brain - like pulse scheme can support robots to work on long - term and complex tasks with low power consumption, enabling robots to get rid of pre - programming, remote controls, and user prompts and achieve full - autonomous work, with strong generality and generalization ability.

Yingke has learned that "Qianjue Technology" has cooperated with Dreame's floor - sweeping robot business and is trying to implement its AGI large model on a new type of floor - sweeping robot, providing the product with a large task - scheduling model and a large visual model for scene understanding. By integrating "Qianjue Technology"'s embodied brain, Dreame's floor - sweeping robot is expected to significantly improve its understanding ability of general scenarios and adaptability to complex scenarios.

Comments from investors:

Gu Shitao, Investment Director of Zhuichuang Venture Capital said that with the continuous expansion of the application scenarios of embodied large models and the rapid improvement of the maturity of robot bodies and core hardware, embodied intelligence is about to reach the inflection point of large - scale commercialization. We are optimistic about the innovative research direction of the Qianjue team in the field of "brain" models. In the future, Dreame will also rely on its own ecosystem to empower the commercialization of Qianjue Technology and jointly expand the application boundaries of embodied intelligence. Zhuichuang Venture Capital will continue to focus on the embodied intelligence track and will continue to deploy in software, hardware, infrastructure, and downstream applications in the future, providing rich scenarios within the ecosystem to boost the emergence and industrial development of embodied intelligence and make technology accessible to thousands of households.

Investment leader of DT Capital said that we are particularly optimistic about the application prospects of large models in the field of robotics. As an incubation result of a national - level research group at Tsinghua University, Qianjue Technology has a rich household scenario dataset. With its unique brain - like model architecture, it effectively improves the data utilization efficiency and successfully breaks through key problems such as the limited decision - making ability and insufficient perception generalization commonly existing in the current robotics field. The company has now established in - depth cooperation relationships with several leading robot enterprises. We firmly believe that Qianjue Technology will continue to promote the intelligent upgrading process of the robot industry and bring more breakthrough innovations to the expansion of embodied intelligence application scenarios.

Investment leader of Jingye Intelligent Industry Fund said that as a new force in the field of embodied intelligence in China, Qianjue Technology focuses on long - term task planning and the brain - like neural network model. It endows robots with stronger generality and autonomous decision - making ability through innovative models and algorithms. With a solid scientific research background, excellent algorithm accumulation, and keen market insight, the Qianjue Technology team has quickly become an important innovative force in this field. We believe that Qianjue Technology will continue to lead the frontier exploration and commercialization of the embodied intelligence brain.