36Kr Exclusive | Former Bosch Autonomous Driving Algorithm Engineer Founds Startup to Develop Tactile Large Model with Synthetic Data

Haptic Glove (Source: Enterprise)

This article contains approximately 2400 words, with an estimated reading time of 6 minutes

Author | Xue Ou

Editor | Silai Yuan

Hard Krypton learned that spatial intelligence firm Dayan Technology has recently completed a multi-million-yuan Angel round financing, led by Songhe Capital, with participation from state-backed institutions including Zhejiang Provincial Financial Holdings and Guangzhou Panyu Innovation Fund. The funds will be primarily allocated to haptic large model R&D, robotics data production line construction, and team expansion.

Founded in May 2025 in Tongxiang, Zhejiang Province, Dayan Technology is a spatial intelligence company focused on 4D world reconstruction, world models, and controllable diffusion model technologies.

Founder and CEO Yang Lin holds a PhD in Computer Science from the National University of Singapore, with 7-8 years of in-depth experience developing autonomous driving algorithms at enterprises including Horizon Robotics, Coreway, BYD, and Bosch Suzhou. Additionally, the company's core technology is derived from over two decades of digital twin and haptic interaction research accumulated by the team of Chief Scientist, Academician of the United Nations Academy of Sciences, and Academician of the Canadian Academy of Engineering, Abdulmotaleb El Saddik.

Yang Lin told Hard Krypton that he observed previous market demand for data was dominated by automotive synthetic data, but starting in 2025, robotics data orders have surged dramatically.

"Automotive data volumes haven't grown, and have even slightly declined. As competition among robot hardware intensifies, the industry is shifting focus to advancing robot intelligence, and the bottleneck for that intelligence lies in data," Yang Lin stated.

Currently, Dayan Technology has established three product lines:

First, synthetic data services: Focused primarily on synthetic data, combined with first-person perspective "heterogeneous acquisition" data collection to support autonomous driving and robot training. The product boasts significant cost advantages—traditional manual annotation of a single frame can cost over ten yuan, while Dayan's synthetic data only costs a few cents per frame, with gross margins exceeding 60%.

Second, haptic data collection equipment: The company has independently developed China's first haptic glove with force-feedback interaction, named "Shadow Gauntlet", alongside a head-mounted device. Featuring 29 array units, 1015 haptic contact points, and a maximum response frequency of 300Hz, these devices can simultaneously capture human hand posture and haptic feedback.

Head-Mounted Device (Source: Enterprise)

Third, haptic large model: This represents the company's ultimate goal, scheduled for release in the second half of this year. The model accepts multimodal inputs, incorporates physical constraints in the latent space, and outputs the force direction, magnitude, and optimal posture when gripping objects. No comparable public product currently exists in China.

After just about one year of operation, Dayan Technology has already achieved commercial implementation. The company's revenue exceeded 10 million yuan in the first quarter of this year, and it has established partnerships with several leading robotics firms, with plans to deliver its first custom embodied intelligence "brain" project by the end of the year.

Furthermore, global expansion is a key priority for the company next year. It plans to establish a subsidiary in Saudi Arabia between late 2025 and early 2026 to drive overseas sales of its hardware and data services. Concurrently, the company does not rule out extending downstream once its haptic large model matures, to develop differentiated "robot intelligence" solutions.

The following is an edited excerpt from Hard Krypton's interview with Yang Lin:

Hard Krypton: What is the biggest technical bottleneck in the current robotics data industry? How are you addressing it?

Yang Lin: There are two main bottlenecks right now. The first is efficiency—the production capacity of the data flywheel cannot keep up with demand, and the iteration speed of domestic autonomous driving models clearly lags behind Tesla. The second is precision—clients have evolved from requiring upper-body posture tracking to full-body tracking, and now need solutions to extract leg postures even under occluded conditions. Pure vision cannot solve the problem of hand occlusion, so haptic data must be integrated.

Each individual module has mature existing technologies, but embedding all of them together while maintaining efficient operation increases complexity exponentially. Our approach follows two parallel paths: first, combining synthetic data with heterogeneous acquisition to rapidly generate large volumes of valid data at extremely low cost; second, independently developing haptic data collection equipment to directly capture haptic information from occluded areas at the hardware level—a practice rarely seen in the industry.

Hard Krypton: Your heterogeneous acquisition cost is far lower than real-device collection cost. How does this business model operate sustainably?

Yang Lin: Our specific method is to partner with scenarios like unmanned convenience stores and front-end warehouses. We have staff wear our head-mounted devices and gloves during their regular work shifts, providing a daily subsidy of 100 yuan to obtain 8 hours of continuous data. Of course, not all 8 hours of data are directly usable—after processing through our data production line, only half or less may remain. Even so, the cost is still an order of magnitude lower than real-device data collection.

Why is real-device collection so expensive? Operators need to build dedicated production facilities, partition spaces into small work cells, and have robots repeatedly perform the same tasks. This costs thousands or even tens of thousands of yuan per day, while the proportion of valid data remains low.

Our logic is: since robots will eventually work like humans, directly collecting human motion data and mapping it to robot configurations—despite some inherent loss—allows massive volumes of human data to compensate for that error. Moreover, our data production line is continuously being optimized, steadily reducing the loss rate over time.

Hard Krypton: How high is the technical barrier for your haptic glove and head-mounted hardware? Could competitors easily replicate them?

Yang Lin: Let me put it this way—no product on the market currently integrates both haptic feedback and hand skeleton posture tracking into a single glove. This involves a series of complex engineering challenges including sensor design, circuit layout, crosstalk prevention, and anti-interference measures, none of which are trivial. The hardware threshold for the glove itself is quite high, and we have filed multiple patents for core processes.

But more critically, the true value lies in the integration of hardware and software. Our devices are purpose-built to serve the haptic large model, with data formats, calibration methods, and preprocessing workflows deeply coupled with model training. Even if competitors disassemble our glove and replicate identical hardware, without our proprietary models and algorithms, the device's utility is drastically diminished. Conversely, our model training relies on this hardware ecosystem to continuously generate high-quality data. This creates a closed-loop system of hardware and software that cannot be replicated through simple, single-dimensional copying.

Investor Perspective:

Songhe Capital stated: Dayan Technology is a spatial intelligence firm specializing in 4D world reconstruction, world models, and controllable diffusion model technologies and applications. It provides high-quality synthetic data and high-fidelity pre-training environments for industries including autonomous driving, embodied intelligence, and digital twins.

Value Proposition 1 – Rare Team Composition: The team is relatively young overall, with core members holding 5-10 years of R&D experience in autonomous driving and AI. Their expertise complements each other across core technical areas including algorithms, perception, and decision systems, alongside product and market capabilities. With long-standing collaborative history, the team demonstrates strong cohesion.

Value Proposition 2 – Significant Market Potential: The synthetic data market is in a phase of rapid growth, with China's market size reaching 2.1 billion yuan in 2024, projected to hit 23 billion yuan by 2030. The embodied intelligence robotics industry is widely recognized as the next trillion-yuan market following autonomous driving.

Value Proposition 3 – High Project Uniqueness: The company's spatial intelligence solution based on world models enables precise perception, modeling, and decision-making for complex environments, making it one of the few domestic startups operating in this specialized field.

Value Proposition 4 – Products Validated by Clients: The offerings have transitioned from lab to market, gaining client validation and industry recognition. The business model has received preliminary verification, laying a solid foundation for large-scale commercial deployment.

Value Proposition 5 – Haptic World Model Fills Domestic Gap: Leveraging Academician Abdulmotaleb El Saddik's leading position and influence in global haptic technology, the company goes beyond embodied data services to deliver end-to-end model layer development and implementation capabilities.

Value Proposition 7 – Elite Government Talent Policy Support: The company's Chief Scientist, Academician Abdulmotaleb El Saddik, has been selected for Zhejiang Province's Kunpeng Talent Program, expected to receive nearly 90 million yuan in various government funding over the next five years. Academician Abdulmotaleb El Saddik (Torch Program) and Professor Dong (Qiming Program) are scheduled to join full-time in May this year, endowing the team with formidable technical strength.