A former senior executive of Yitu Technology raised tens of millions of yuan in startup financing, bridging the physical world to AI models and promoting the intelligent transformation of devices | Exclusive from 36Kr

Author | Zhang Ziyi

Editor | Peng Xiaoqiu

Yingke has learned that Ningbo Yunjin Micro Intelligent Technology Co., Ltd. (hereinafter referred to as "Yunjin Micro") recently completed its Angel + round of financing, raising 10 million yuan. The investment institutions in this round are Shanghai Hengdi and Ningbo Shungong. The financing funds will be used to iterate and upgrade the platform, expand equipment products, and increase the ecological coverage of vertical industry scenarios.

Founded in June 2021, Yunjin Micro is a technology company focusing on the research and development of embodied intelligent agent operating systems. The company is committed to developing an AI brain for embodied intelligence, helping enterprises apply large - model technology to edge computing, and realizing low - cost intelligent transformation of systems and equipment.

Currently, there is a trend for large models to move from the cloud to edge computing, which is expected to open up a trillion - level market. Research data shows that the global market size of embodied intelligent devices exceeded 25 billion US dollars in 2024, with a compound annual growth rate of nearly 20%. In China, the demand for intelligent transformation in areas such as industrial automation and smart cities accounts for more than 35%.

Yunjin Micro has noticed the urgent need for multi - modal AI in the physical world. Traditional large language models can only process one - dimensional text data, while scenarios in industries such as industry, retail, and security require "embodied intelligence" that integrates multi - dimensional data such as vision, sensors, and control instructions. Its independently developed Yunjin OS operating system can reconstruct the interaction logic between AI and the physical world through two core technologies.

Specifically, Yunjin OS's unique MaM (Model - Alloy - Model) model synthesis model realizes nanosecond - level collaborative scheduling of multiple heterogeneous models such as vision, sensors, and control instructions through routing hybrid technology. Taking the intelligent inspection scenario as an example, the vision model is responsible for identifying equipment defects, the sensor model analyzes temperature and humidity data, and the control model automatically triggers alarm instructions. The three achieve efficient collaboration through MaM, with the efficiency several times higher than that of single - model solutions.

More importantly, this architecture solves the training problem caused by data fragmentation in the physical world - more than 90% of private multi - modal data does not need to be uploaded to the cloud and can directly complete the closed - loop of "data collection - model fine - tuning - instruction execution" on edge devices, significantly reducing data security costs.

In response to the heterogeneous characteristics of domestic computing power, the VT - Transformer minimalist computing framework developed by Yunjin Micro uses 30,000 lines of C++ code to achieve the functions of 600,000 lines of Python code in traditional AI frameworks, reducing model inference latency to 12 ms and reducing video memory usage by 85%.

The above technological breakthroughs enable multi - modal models with tens of billions of parameters to run on edge hardware at a cost of only a few hundred yuan, significantly reducing computing power costs and equipping devices such as intelligent cameras and industrial controllers with more cost - effective "large - model brains".

Currently, through the "low - cost intelligent transformation" model, Yunjin OS has successfully entered scenarios in energy, manufacturing, retail, etc. As of Q2 2025, the company has served more than 120 enterprises, achieving tens of millions of yuan in revenue. Among its customers are large technology and industrial enterprises such as China Electronics, Guiyang Rail Transit, SAIC Group, and Shanghai Tunnel Engineering.

Wang Wenyi, the founder of Yunjin Micro, believes that the company will adhere to the vision of "making AI affordable for every enterprise", help enterprises complete the training and inference calculations required for intelligent agents at extremely low costs, and promote the inclusive development of AI technology.

In terms of the team, Wang Wenyi graduated from Zhejiang University with a bachelor's, master's, and doctor's degree. He once managed the hardware division of Yitu Technology, one of the "Four Little Dragons of CV", and led the implementation of several industry - benchmarking AI hardware projects. Co - founder Dr. Zhou Chang also graduated from Zhejiang University with a bachelor's, master's, and doctor's degree. He once served as the R & D leader of the large - scale vision model for the City Brain at Alibaba DAMO Academy and has rich experience in the field of multi - modal algorithms.

The team members include those with more than 10 years of experience in fields such as system software, chip design, and visual AI. The company has established a "production - learning - research integration" cooperation model with scientific research institutions such as the Shanghai Advanced Institute of Zhejiang University, the Robotics Research Institute of Zhejiang University, and the Knowledge Graph Laboratory of Zhejiang University, giving Yunjin Micro technological advantages in core areas such as miniaturization of multi - modal large models, edge computing power scheduling, and incubation of embodied devices.