Enable robots to have a sense of touch. Xinzhi Embodied completed nearly 100 million yuan in angel round financing.

Investment community AI has learned that today (May 27), Shanghai NeoteAI Embodied Intelligence Technology Co., Ltd. (NeoteAI for short) announced the completion of an angel round of financing of nearly 100 million yuan. This round was jointly led by Shanghai Science and Technology Innovation Group under Shanghai State Investment and Fudan Science and Technology Innovation, and co - invested by Ceyuan Fund under Shanghai Science and Technology Innovation Group. Duowei Capital served as the exclusive financial advisor.

This is a Fudan - affiliated team. CEO Zhao Shihao graduated from Fudan University. The team's chief scientist Wu Zuxuan and COO Dong Daoguo are the deputy dean and researcher of the Institute of Trustworthy Embodied Intelligence at Fudan University respectively. Different from most enterprises in the industry, NeoteAI chooses to use touch as the primary modality of embodied intelligence and has a full - stack layout from tactile sensors, tactile data to tactile large models.

To enable machines to truly integrate into the physical world, having only vision is far from enough. Currently, tactile data is becoming an indispensable part of the robot model training chain. When embodied intelligence starts to move towards real - world scenarios, this also becomes the key breakthrough for industrialization.

Born out of Fudan

They "Add Touch" to the Embodied Large Model

The NeoteAI team was born out of the Institute of Trustworthy Embodied Intelligence at Fudan University.

CEO Zhao Shihao was born in 1997. He graduated from Fudan University for his undergraduate and master's degrees and is currently pursuing a doctorate at the University of Hong Kong. Before starting his business, he was a core researcher at Microsoft Research and Alibaba Tongyi Lab, focusing on the research and development of cutting - edge models, covering video world models and generative models.

Zhao Shihao, Wu Zuxuan, Dong Daoguo

Among the team members, Chief Scientist Wu Zuxuan is the deputy dean of the Institute of Trustworthy Embodied Intelligence at Fudan University and has worked at Meta. He has long been engaged in research on video models and multi - modal models. COO Dong Daoguo is a cross - border talent with both academic and industrial experience. He has nearly 20 years of experience in the industrial field. He was the chief architect of Huawei Honor Magic 1 and is currently a researcher at the Institute of Trustworthy Embodied Intelligence at Fudan University, mainly ensuring the commercialization of the company's technology.

In June 2025, Shanghai NeoteAI Embodied Intelligence Technology Co., Ltd. was officially established in Shanghai, positioning itself as an embodied intelligence model company.

In their view, vision allows robots to see the world, but touch allows robots to truly perceive the physical world. As robots enter various complex scenarios, they are bound to come into contact and interact with the physical world. Only touch can enable robots to truly perceive contact, understand materials, and judge forces, and complete stable operations after accurate capture. In short, the introduction of the tactile modality is a key step for embodied intelligence to move from "understanding tasks" to "completing tasks".

For touch to truly empower embodied intelligence, it is far more than just installing a sensor. NeoteAI is building a complete technical system for the real physical world, including upstream sensors, mid - stream data collection, and downstream models. Specifically:

Sensor layer: Self - developed visual - tactile sensor. For grippers and dexterous hands, it uses a monochromatic light + particle scheme and model decoupling technology to convert force, slip, deformation, and boundary information during contact into high - density visual representations. The modular design enables 24 - hour high - intensity continuous use. It is reported that this sensor was led by researcher Chen Wenming and his doctoral student Luo Hu from the Institute of Trustworthy Embodied Intelligence at Fudan University, which is also the first landmark achievement of the company's industry - university - research integration with Fudan University.

(NeoteAI Visual - Tactile Sensor)

Data layer: The "fuel factory" for tactile models. Currently, NeoteAI has built a professional tactile data collection factory covering an area of over 1,000 square meters, covering a large number of refined operation tasks such as industrial assembly, wire harness operation, and flexible object handling. The collection of tactile data has extremely high requirements for equipment. Each end - effector needs to be equipped with a tactile sensor perception module, which is coordinated with a force feedback system, synchronous collection, and subsequent annotation. The collected data includes not only image sequences but also physical information such as contact force, slip, deformation, and texture.

Model layer: An embodied large model containing the tactile modality. Based on large - scale systematic tactile embodied data, NeoteAI is developing an embodied large model containing the tactile modality, aiming to integrate the tactile modality into the mainstream embodied framework. It has now achieved a major breakthrough in fine operation tasks, with accuracy, stability, and generalization ability leading the industry.

Traditional VLA relies on vision and language and is prone to failure due to perceptual blind spots. NeoteAI's VTLA model can obtain real - time contact feedback such as clamping, slipping, and deformation to accurately guide operations. Its tactile world model makes up for the shortcomings of vision, with the prediction success rate in refined scenarios increased by more than 90%. In reinforcement learning, NeoteAI uses touch as the "golden signal" for strategy optimization to correct operations in real - time, reducing the error rate of high - difficulty fine tasks to nearly zero.

Start Verification from Factory Tasks

Move towards Commercialization

Now, NeoteAI has successfully received its first round of financing.

In fact, as an important implementation result of the strategic cooperation between Fudan University and Jing'an District, NeoteAI received strong support from the Jing'an District Science and Economic Commission and Shibei High - tech Group during its start - up period. It has successively received strategic funds from Jing'an District and special funds from the Shanghai Municipal Economic and Information Commission to promote high - quality industrial development. The Shanghai Science and Technology Commission also provided special funds for the research and development of its core visual - tactile sensor, comprehensively accelerating the company's technological research process.

There is still a long way to go from the tactile embodied model to general robots, but in the short term, a clearer commercial entry point lies in factory and semi - structured scenarios.

Currently, NeoteAI's visual - tactile sensors have entered the customer system of leading embodied intelligence companies and are being verified in scenarios such as fine operation, data collection, and model training. More importantly, its developed tactile model has been piloted in industrial fine operation scenarios and has won multiple POC orders.

The team said: "The difficulty of fine operation is not only to enable robots to see but also to make correct judgments when contact occurs. NeoteAI is simultaneously deploying sensors, data platforms, and models, hoping to transform real and continuous tactile data into the learnable and generalizable abilities of robots, ultimately improving their autonomous perception, motion planning, and task execution levels."

Industry analysis believes that in the next 1 - 2 years, touch will become a standard perception ability for humanoid robots, industrial robots, and service robots. The integration of the "vision + touch" dual - modality is expected to become the mainstream trend in the industry. In other words, the one who can truly enable robots to complete fine operation tasks across scenarios and across entities will be closer to the core of the current round of embodied intelligence competition.

As the leading investor in this round, Shanghai Science and Technology Innovation Group said that the company has always been concerned about hard - tech enterprises with underlying technological innovation capabilities, the ability to solve real industrial pain points, and the potential to become the cornerstone of future industries. "We are optimistic about NeoteAI. The core lies in the team's promotion of tactile perception, data collection, and model training in the same technological closed - loop. As embodied intelligence moves towards more complex real - world scenarios, from industrial assembly to changing light bulbs, cutting tofu, and using scissors in household services, tactile data will become a key variable in improving the fine operation ability of robots."

After the completion of this round of financing, NeoteAI will continue to promote the mass production and delivery of sensors, expand the scale of tactile data collection, and accelerate the training and verification of the VTLA model, tactile world model, and tactile reinforcement learning in real - world operation tasks, promoting the tactile embodied technology to move from the laboratory to real - world scenarios such as factories and households.

This day is approaching at an accelerating pace.

This article is from the WeChat official account "Investment Community AI", author: Yang Jiyun. Republished by 36Kr with permission.