Google DeepMind director behind Gemini leaves for Hong Kong university after 20 years in US

Cao Liangliang, a former Google DeepMind director who worked on Gemini, has left the U.S. to join Hong Kong Polytechnic University, two decades after he left the city to build his career abroad.

PolyU appointed Cao Chair Professor of Artificial Intelligence Systems in its Department of Data Science and Artificial Intelligence, effective June 29, the university announced, as reported by the South China Morning Post.

"Returning to Hong Kong is a full-circle moment for me," Cao wrote on his personal website.

His path in AI began at the Chinese University of Hong Kong in 2003, in the Multimedia Laboratory run by Tang Xiao'ou, the computer scientist who went on to found the AI company SenseTime.

Tang, an MIT-trained researcher who joined CUHK's Department of Information Engineering in 1998, turned the lab into a training ground for many of the people now shaping China's AI industry. He died in December 2023, Caixin reported.

Cao entered the University of Science and Technology of China in 1998 and earned his bachelor's degree in 2003 before joining Tang's lab, where he completed a master's in 2005.

In 2006 he moved to the U.S. for a doctorate at the University of Illinois Urbana-Champaign under Thomas S. Huang, a pioneer of computer vision who mentored more than 100 students before his death in 2020.

Cao Liangliang, a former Google DeepMind director who worked on Gemini. Photo courtesy of Cao Liangliang

In 2010 Cao was part of a team that won the ImageNet Large Scale Visual Recognition Challenge, the annual contest built on the dataset created by Stanford computer scientist Li Fei-Fei.

The challenge became a foundation of the deep-learning breakthroughs that followed in computer vision.

After his doctorate, Cao went into industry research, joining IBM's T.J. Watson Research Center in 2011 and later Yahoo Labs, while teaching as an adjunct professor at Columbia University and the University of Massachusetts Amherst.

He co-founded the start-up Switi Inc., which Google acquired in 2018, then worked at Google as a senior staff scientist and manager.

He moved to Apple as a principal scientist and engineering lead, serving as the modeling lead for on-device Apple Intelligence, before returning to Google, now Google DeepMind, as a principal engineer and director on the Gemini team.

"For more than a decade, I have worked on intelligent assistants and AI agents, contributing to Gemini, Google Assistant, Apple Intelligence and IBM Watson," Cao wrote on his website.

His recent research centers on AI tools for special education, including methods to detect early signs of autism by analyzing a child's gaze and facial expressions in video.