Recruiting

We are Recruiting Ph.D. students, Master students, Postdocs, and Undergraduate Research Interns. Contact us if you are interested in Embodied AI.

招募 博士后、博士、硕士以及本科实习生(详情)，欢迎对智能机器人感兴趣的同学加入我们！

Short Bio

Cewu Lu, Professor at Shanghai Jiao Tong University, Distinguished Professor of the Changjiang Scholars Program (长江学者特聘教授), and recipient of the Scientific Exploration Award. In 2016, he was selected as a high-level overseas young talent, and in 2018, he was named one of the 35 Innovators Under 35 in China by MIT Technology Review (MIT TR35). In 2019, he received the Qiushi Outstanding Young Scholar Award, and in 2020, he was the third contributor to the Shanghai Science and Technology Progress Special Award. In 2022, he received the Ministry of Education Youth Science Award and was recognized for one of the six best papers at IROS (out of 3579 submissions). In 2023, he was nominated for the Best System Paper Award at the Robotics: Science and Systems (RSS) conference (one of four nominations) and received the Scientific Exploration Award (the only recipient in the field of embodied intelligence).

Before he joined SJTU, he was a research fellow at Stanford University working under Prof. Fei-Fei Li and Prof. Leonidas J. Guibas. He was a Research Assistant Professor at Hong Kong University of Science and Technology with Prof. Chi Keung Tang. He got the his PhD degree from The Chinese Univeristy of Hong Kong, supervised by Prof. Jiaya Jia.

As a corresponding author or first author, he has published more than 100 papers in high-impact journals and conferences, including Nature, Nature Machine Intelligence, and TPAMI. He serves as a reviewer for Science, Nature sub-journals, Cell sub-journals, and as area chair for NeurIPS, CVPR, ICCV, ECCV, IROS, and ICRA. His research interests include embodied intelligence and computer vision.

Our Vision

The development of Robots for general-purpose has long been a shared dream of humanity, as their realization would significantly enhance productivity—by, for instance, performing tasks typically undertaken by nurses or cleaning staff—and improve the quality of life through applications such as domestic service robots. A general-purpose robot must be capable of executing a wide range of tasks in diverse and open-ended environments, posing a formidable challenge in the field of artificial intelligence. The crux of this challenge lies in enabling robots to acquire human behavioral capabilities. Building upon a strong foundation in the visual understanding of human behavior, we aim to explore a novel approach: empowering robots to learn comprehensive, general-purpose behaviors by observing and interpreting vast amounts of human activity in video form. Compared to the mainstream approach of guiding robotic behavior through large language models, our strategy offers several advantages in achieving generalizability:

• Autonomous Learning: Human behavior videos capture the full scope of tasks and their various possibilities, circumventing the limitations of manually defined tasks.


• Scalability: The extensive accumulation of videos through the growth of the internet covers nearly every conceivable human task and operation.


• Automatic Expansion: With large volumes of new videos uploaded daily, robots can automatically extend their capabilities to include new tasks and skills.

By implementing this innovative approach, we pave the way for a more universally applicable and effective solution to the challenge of creating general-purpose robots.