Shanghai Jiao Tong University's Entrepreneurial Army Gains Momentum, Yitu Technology Secures Investment from Xuhui Capital, Seizing the "Deep Water" of Automotive Models

Amid the bustling Shanghai Auto Show, a reporter from the Science and Technology Innovation Board Daily learned that Shanghai Yitu Technology Co., Ltd. (hereinafter referred to as "Yitu Technology"), which develops intelligent automotive software, has also entered the field of vertical automotive models.

Pan Li, CMO of Yitu Technology, told the reporter that the company, based on its industry know-how over the past decade and its grasp of AGI technology development, sees the potential in vertical automotive models and AI systems, thus focusing on this field."The newly launched Xinjie AI full-stack system makes the automotive AI brain smarter, with deeper understanding and more advanced reasoning capabilities."

Yitu Technology was established in June 2023 and recently completed a Pre-A round of financing worth tens of millions, led by Xuhui Capital, with participation from Dening Capital and Yunqi Capital. The funds will be primarily invested in the R&D of vertical automotive models and the intelligent space AIOS system, promoting mass production and batch delivery.

Shanghai Jiao Tong University Alumni Co-Found "Automotive AGI"

As an R&D company in the vertical field of automotive AGI, the reporter noted that the founders of Yitu Technology are all from Shanghai Jiao Tong University.Among them, CEO Xiaohang Wu graduated from Shanghai Jiao Tong University and has 10 years of experience in the strategic commercialization of the intelligent automotive industry; COO Yongfeng Xia also graduated from the same university and has 14 years of experience in intelligent automotive R&D.

Based on the founders' deep experience in the automotive industry, Yitu Technology has more insights into the development of vertical automotive AGI.

Pan Li told the reporter that with the rapid evolution of foundational models, such as large language models (LLMs), voice models, multimodal models, and spatial models,the definition and development methods of automotive products are undergoing fundamental changes, shifting from a hardware and software-dominated approach to a model-defined, data-driven paradigm.

In this context, cars are no longer just means of transportation but are evolving into AI companions with emotional interaction capabilities and the ability to achieve service closed loops. This has become an industry consensus.

It is reported that in the exploration of AGI in the automotive field, Yitu Technology aims to deepen its technical expertise and gain market insights to find commercial opportunities. "Currently, Yitu Technology has completed product R&D and business implementation from 0 to 1. At its inception, the company launched an in-vehicle intelligent platform in the vertical industry and successfully partnered with several OEMs for mass production and installation. Now, with the continuous innovation of foundational models, Yitu Technology has introduced the Xinjie AI full-stack system with deep reasoning capabilities."

Regarding this AI full-stack system, Pan Li explained:It is a native AI Native system for the AI automotive era, including Xinjie AI-Mind, Xinjie AI-HMI, and Xinjie AIOS, embodying Yitu Technology's new understanding of automotive intelligent space AI architecture, AI scenarios, and AI interactions. The development of this system took nearly a year.

Pan Li believes that this AI full-stack system can deeply understand user needs in automotive scenarios. "For example, automatically dimming lights, playing soothing music, recommending nearby cafes and completing orders, achieving a complete service chain closed loop; and when a user says 'I'm hungry,' the system will consider multiple factors such as current location, traffic conditions, and user dietary preferences to provide the optimal solution."

"In the intelligent era, users need to free their hands during car travel, and automotive AGI is the best carrier," Pan Li said.

Under the DeepSeek moment, three key periods in the development of automotive AGI cannot be ignored.

Yitu Technology's exploration is a microcosm of automotive AGI development. Looking at the entire industry, three key aspects are crucial for achieving AGI in the automotive field.

"At the foundational model level, its development determines the construction of the upper ecosystem."An investor told the reporter that currently, domestic producers such as DeepSeek, Doubao, and Tongyi Qianwen are leading the way. The product release pace and technical direction of these three in the next six months are key to understanding industry dynamics. "Taking DeepSeek as an example, its reasoning capabilities have become an industry benchmark, and AI agents developed around reasoning capabilities are rapidly advancing."

Secondly, chips remain crucial."Although DeepSeek has changed the demand for computing power through algorithms, with the rise of complex models like multimodal models, efficient edge-side deployment has become essential. Currently, the iteration cycles of chips from NVIDIA, Qualcomm, and MediaTek (MTK) have shortened from two years to just over a year. Rapid technological updates provide hardware support for model deployment, and only by adapting to advanced chips can foundational models perform optimally."

Finally, the operating system.As the most fundamental infrastructure, the operating system is the foundation for AI system operation. Today's advanced AI systems must rely on foundational operating systems to run stably. "These are the three most important elements for the industry's development, competition, and transformation. Only with the development of these three elements will AGI in the automotive field and the entire AGI move forward."

Returning to the development of automotive AGI, Pan Li believes that automotive AI agents are just the first step. To turn cars into AGI terminals, autonomous driving (AD) needs to reach L4 level, and the perception, understanding, and service capabilities within the cabin, as well as the intelligent scheduling and coordinated control of chassis, suspension, and other vehicle parts, should also reach L4 autonomous level.

"Recently, autonomous driving companies have been trying to integrate language models into visual perception (VLA). From the perspective of intelligent space development, AD focuses on driving environment and vehicle state perception, while intelligent space covers language, images, and vehicle signals. The two are likely to merge technologically, and this fusion might happen within three years."Pan Li said.