The 2025 MIT Technology Review Innovators Under 35 (TR35) China Region list was released, and three alumni from Huazhong University of Science and Technology (HUST) have been selected. The honorees are Li Boyi, who enrolled in 2015 at the School of Computer Science and Technology, and Zhao Hengshuang and Ding Wenchao, who enrolled in 2011 at the School of Electronic Information and Communications.

The TR35 list was created in 1999 and introduced to China in 2017. It focuses on recognizing young researchers who have made outstanding achievements in fundamental research, breakthroughs in core technologies, and the commercialization of research results. To date, the TR35 China list has highlighted 315 young scholars, facilitated over 100 groundbreaking scientific discoveries, and incubated nearly 100 high-potential technology companies.

Li Boyi (Senior Research Scientist at NVIDIA and University of California, Berkeley)

Li Boyi specializes in multimodal vision and language understanding integration technology. Before pursuing her PhD, she pioneered an end-to-end neural network to replace traditional image dehazing algorithms, which became a benchmark in the field. During her doctoral studies, she developed LSeg, introducing open-vocabulary visual recognition technology.

After joining University of California, Berkeley (UC Berkeley) and NVIDIA, her research has been applied to autonomous driving and service robots. She has built object-level multimodal models to handle unknown road scenarios and established a framework for interactive planning with large models, enabling robots to execute voice commands. She continues her work in developing intelligent caregiving systems for home use.

Zhao Hengshuang (Assistant Professor at University of Hong Kong)

Zhao Hengshuang’s research focuses on physical intelligence. In his early work, he proposed PSPNet, a foundational algorithm for semantic segmentation, widely applied in autonomous driving and medical image detection. Zhao Hengshuang also developed the Point Transformer series of algorithms and Depth Anything series of models to enhance 3D spatial perception.

He later introduced DriveGPT4 and GPT4Point, combining spatial perception with large language models. This work has established a comprehensive 2D, 3D, and multimodal intelligent technology framework, applicable to autonomous driving and general-purpose robotics.

Ding Wenchao (Early-Career Researcher at Fudan University and Co-founder and chief scientist of TARS Co)

Ding Wenchao focuses on decision-making algorithms for agents operating in complex environments. He was one of the first to propose an autonomous driving decision-making framework that integrates environmental prediction and contributed to building Huawei's ADS system, which has been mass-produced in over a million vehicles.

After launching his own venture, Ding Wenchao proposed a human-centered data route and the AI world engine embodied foundation model. His approach uses human behavior data to train robots, reducing reliance on rule-based programming and advancing the commercialization of general-purpose robotics.

Source: Office of Alumni Affairs and Office of the Education Development Fund, HUST