The former lead of the world foundation model from Huawei's Automotive BU has founded a new venture, and the Hangzhou-headquartered embodied brain company has closed a nearly 100 million yuan financing round | HardKrypton Exclusive.
Author | Huang Nan
Editor | Yuan Silai
36Kr has learned that embodied intelligence startup GeWuTongZhi has completed a nearly 100 million-yuan angel round of financing. Investors include Peak Investment, Hongfu Assets, Hangzhou Capital, Hangzhou Talent Group, a provincial-level embodied industry investment institution, and many leading listed companies. The raised funds will be mainly used for large-scale pre-training and technical iteration of the physical inversion world model, hardware and ontology construction, accelerating talent introduction and team expansion, and promoting PoC verification in multiple scenarios and real data collection. Maple Pledge Capital acted as the exclusive private equity financing advisor for this round.
Headquartered in Hangzhou, GeWuTongZhi was established in August 2026. It received support from three Hangzhou municipal-level funds in its first round of financing, including Hangzhou Runmiao Fund, Hangzhou Disruptive Achievement Transformation Fund, and Hangzhou Talent Fund. It is an ecological partner of the national pilot base for embodied intelligence application and a member of the Hangchuang Embodied Intelligence Camp.
The company develops self-evolving dexterous operation intelligence for real industrial scenarios. Relying on the real closed loop of multi-modal data dominated by haptics and vision, it builds a technical system of "physical inversion world model + PSI physical self-evolution", providing robots with an embodied brain with "haptic perception", which can independently accumulate practical experience in continuous operations, convert "working seniority" into capabilities, and realize continuous evolution that becomes more adaptive to scenarios as it is used.
Among the investors of this round, Peak Investment, as a leading domestic sports technology industrial group, has signed a strategic cooperation agreement with GeWuTongZhi. The company will focus on the requirements of "double flexibility + one high precision" in footwear manufacturing, namely flexible manufacturing, flexible materials, and sub-millimeter precision, to obtain large-scale orders in the initial stage of the enterprise, and jointly promote technology implementation and industrial verification.
Dr. Cai Jia, Founder and CEO, graduated from the School of Astronautics, Northwestern Polytechnical University, and studied under Professor Huang Panfeng, a recipient of the National Science Fund for Distinguished Young Scholars, focusing on the research of on-orbit dexterous operation of space robots. Later, he joined Huawei 2012 Laboratory and the Automotive BU 2030 Laboratory, deeply participated in the core technology R&D and million-level mass production delivery of ADS 2.0 to 4.0, as well as the pre-research of ADS 6.0, served as the director of a 100-person team and a member of Huawei Scientist Committee.
He also has threefold compound backgrounds in space operation robots, autonomous driving mass production, and world models, with engineering experience in cutting-edge AI technology incubation, key technological breakthroughs of complex intelligent systems and million-level mass production, and a systematic methodology of ten million-hour level data governance and nearly ten thousand card cluster training. He is one of the core participants of Huawei's intelligent driving system from technical verification to mass production delivery, and has won Huawei's highest personal technical award.
Dr. Cai Jia, Founder and CEO of GeWuTongZhi (Source / Enterprise)
In addition, many core members are from leading enterprises in the industry such as Huawei, Li Auto, Horizon Robotics, and Unitree, and graduated from top universities such as Tsinghua University, Zhejiang University, Shanghai Jiao Tong University, University of Science and Technology of China, Fudan University, Harbin Institute of Technology, Northwestern Polytechnical University and Nanyang Technological University, covering the whole chain of cutting-edge AI research, complex system engineering and large-scale delivery. It is a team with academic research background and mature industrial implementation capabilities.
In the past two years, embodied intelligence has made substantial progress in technical links such as visual perception and motion control, enabling robots to recognize objects, plan paths, and reproduce complex limb movements. However, most of these capabilities remain at the level of long-distance, non-contact perception and execution, and it is difficult to deal with fine operation problems. Once entering the real production site, facing physical interactions at the contact level such as fabric slippage, part jamming, and insertion resistance, relying only on vision cannot obtain the magnitude, direction and contact state of force, and the shortcomings of perception dimension will be directly exposed.
Another underlying bottleneck lies in data. Most embodied intelligence enterprises rely on crowdsourced collection or material library data to train models. Such data has structural deviations from real physical working conditions, and conditions such as illumination, process parameters, equipment anomalies, and human intervention are disconnected from the real environment. More critically, the data in real scenarios has a clear causal chain: "what action leads to what result, what anomaly triggers what adjustment. These causal correlations can hardly be obtained from material data detached from scenarios. Intelligence generated out of scenarios is difficult to be truly implemented." said Dr. Cai Jia, Founder and CEO of GeWuTongZhi.
In the view of the GeWuTongZhi team, it is only the starting point for robots to enter factories. The real challenge is whether it can continue to work on the production line, adjust itself after encountering problems, and bring the experience of this time to the next station.
Cai Jia split this path into three links: Deploy — first put the robot at the customer site to complete a specific process for the customer; Learn — convert the successful experience, failure trajectory, anomaly recovery and manual correction generated in the operation into training data; Generalize — migrate the operation capabilities accumulated from specific projects to the next station, the next task, and the next ontology.
To understand the technical foundation of GeWuTongZhi, we need to return to the core problem of dexterous operation: how robots form "haptic perception".
Traditional robots learn the joint trajectories of human hands. For example, when screwing a bottle cap, the robot will imitate the corresponding action by learning the movement trajectory of human fingers. "But this is only 'similar in form'." Cai Jia told 36Kr, "It does not know how much force people need to use, what the friction coefficient is, and whether to increase the force or adjust the angle when encountering resistance. Once the bottle cap of different materials or the thread with different tightness is replaced, the model may fail."
Aiming at this blind spot, the GeWuTongZhi team first proposed the technical route of "physical inversion world model", which allows robots to infer hidden physical information such as contact and force from visible actions for large-scale pre-training, and then calibrate with real force and haptic feedback to supplement learnable physical labels for action data.
Haptic data is the core of this path and also a recognized problem in the industry. Since the dimension of haptic signals is low, usually only dozens of dimensions, while an image has millions of dimensions, it is extremely difficult to integrate the two in the same model. At the same time, the collection of haptic data is high in cost and slow in speed, and large-scale application has always been a bottleneck.
The solution of GeWuTongZhi is Real to Sim (real to simulation), which uses real physical interaction data to build a simulation environment in reverse, and has determined a series of operation paths.
The core value of this data pipeline is to convert the massive first-person and third-person videos on the Internet into effective data that can be used for model training. Large-scale collection of haptic signals has long been restricted by high cost and low efficiency, while visual data is naturally easy to obtain. Relying on the Real to Sim solution, the team can reversely deduce and reconstruct the corresponding haptic signals from the visual pictures, so that the haptic data can achieve the same collection capacity as the visual data.
At the same time, at the level of world model architecture, the solution greatly improves the signal-to-noise ratio of physical information such as force and haptics, realizes the native integration with visual information, so as to acquire operation haptics beyond human level. Through modeling and predicting physical causality, the model can understand the underlying laws behind the operation, and further strengthen the cross-scenario generalization ability.
On this basis, GeWuTongZhi further builds a PSI physical self-evolution system, which can realize the capability iteration closed loop of "physical inversion - planning execution - real feedback - evaluation learning - verification deployment". The successful experience, failure trajectory and manual correction generated by the robot during operation are filtered and evaluated to improve the physical state estimation and operation strategy, and the updated model is verified before entering the next round of tasks.
Build the technical system of "physical inversion world model + PSI physical self-evolution" (Source / Enterprise)
Cai Jia told 36Kr that the core of this mechanism is two-way improvement: a more accurate world model helps robots make more effective plans, and more valuable execution feedback helps the model identify its own errors and optimize subsequent learning. At the same time, the system can optimize data screening and improve the efficiency of learning from experience through the model's judgment of failure causes and uncertainty.
This has direct value in industrial scenarios. "When traditional robots are deployed to new stations, repeated teaching, debugging and trial and error are required. What we hope to achieve is that the experience accumulated on one site can reduce the migration cost of the next site. When customers add one application scenario each time, what we gain is not only business opportunities, but also reusable deployment experience." Cai Jia said.
36Kr has learned that GeWuTongZhi is promoting joint innovation PoC with customers in the fields of textile and apparel, optical modules and auto parts, and it is expected to complete the first PoC delivery by the end of 2026. The company chooses to cut in from tasks with clear boundaries and quantifiable acceptance standards, gradually expand the coverage of environments and tasks, first break through the Y-axis accuracy success rate and then break through the X-axis generalization. After verifying the basic capabilities in real scenarios, the accumulated experience will be migrated to a wider range of tasks and ontologies.
The following is the excerpt of the interview between 36Kr and Dr. Cai Jia, Founder and CEO of GeWuTongZhi (slightly edited):
36Kr: Haptic data collection has long faced the bottleneck of cost and efficiency. Why did GeWuTongZhi choose the Real to Sim technical path?
Cai Jia: Haptic perception is the core for robots to realize dexterous operation, but haptic data collection itself has always been an industry problem. If we only rely on haptic gloves for collection, not only the cost is high and the collection efficiency is low, but also after the operators wear the gloves, it will even affect their normal daily work. This results in that the amount of data that the gloves can obtain is very limited, which is difficult to support the generalization ability required for embodied model training.
The idea of GeWuTongZhi choosing Real to Sim is very clear: video data is the easiest to obtain on a large scale, so we reversely reconstruct haptic signals such as interactive contact from videos. The accuracy of the haptic data reconstructed by the solution is expected to meet the pre-training requirements, and then the accuracy will be further aligned through real on-site data. More importantly, this pipeline enables haptic data to have the same large-scale potential as visual data. The huge stock of first-person and third-person videos on the Internet can be processed through this process and converted into training samples with haptic labels.
36Kr: You emphasize that quality is better than quantity in data governance. How did this judgment come into being?
Cai Jia: The core lies in whether a set of quantifiable data value evaluation system can be built to automatically filter out low-value redundant data.
This evaluation system is not a simple manual sampling inspection, but scores data from multiple dimensions, such as the rarity of scenarios, the diversity of actions, the proportion of failure cases, the matching degree with the weak links of the current model, etc. Low-score data is directly discarded, and high-score data enters the training set. At the same time, after each round of training, a fixed evaluation set is used to test the performance of the model, find out new weak links, and then reversely guide the next round of data screening.
We can take autonomous driving as an example. If uniform random sampling is adopted, more than 90% of the time during human driving is in a steady driving state, and there are almost no scenarios such as emergency avoidance and complex lane changing. There is no need to accumulate a large amount of such conventional data, and excessive stacking will instead weaken the learning effect of the model in key scenarios. In the past, when we trained the intelligent driving world model, the total data volume of 10 million clips did not change much, but its recipe was optimized and refreshed every week. Therefore, it is necessary to score the data and retain high quality and good diversity samples.
We migrated this idea to the development of the embodied world model. From the first day, the team adhered to the data concept of quality first, and controlled the data redundancy at a very low level. Many peers are competing for the data scale of millions of hours, but a large amount of them are redundant data, and the proportion of truly effective samples is not high. In an environment where storage costs continue to rise, this strategy has advantages. We do not need to consume resources for invalid data.
36Kr: GeWuTongZhi chooses to cut in from industrial scenarios. Different customers have significantly different index requirements for the embodied operation capability of robots. How does the team define the priority of scenario entry?
Cai Jia: Different factories do have significantly different requirements for capability boundaries. The accuracy requirements range from plus or minus 0.1 mm to one or two millimeters, the cycle time ranges from millisecond level to minute level, the success rate requirements range from 90% to 99%, and the reliable working time ranges from 8 hours to 24 hours.
The strategy of GeWuTongZhi is to choose scenarios that match the current capabilities to cut in, and gradually extend to higher requirements. I think the value of a startup lies in identifying and focusing on truly valuable and scalable scenarios, and then expanding the scenarios in a gradual way. Perhaps the "GPT 3 moment" of embodied intelligence will be reached in the process of development.
For specific scenarios, we mainly look at several dimensions: first, the quantifiable degree of the task, whether the acceptance standard is clear; second, the matching degree with the current model capability, whether the accuracy success rate can meet the basic requirements of customers; third, the replicability of the scenario, whether the experience accumulated on one site can be migrated to the next site. At the current stage, three points depend on the model and seven points depend on engineering. For embodied intelligence to be truly practical, it should be generated in this scenario, not out of the scenario in the collected material data.
The indicator we value most is not the accuracy success rate, but the customer repurchase rate. No matter how high the Demo accuracy success rate is, if customers are not willing to repurchase, it means that the product does not really solve the problem. Repurchase means that customers think the input ROI is cost-effective, and means that the robot is truly integrated into the production process.