The person in charge of Apple's robotics has also been poached by Mark Zuckerberg. An alumnus of Zhejiang University, he now holds the top position in Meta's robotics technology.
Apple has lost another four AI researchers, three of whom are of Chinese descent.
Mark Gurman, a journalist who has been following Apple for a long time, reported that Jian Zhang, the chief AI researcher of Apple's Robotics research group, will switch to Meta, and the news has been confirmed by Meta.
Besides Jian Zhang, whose departure is confirmed, Gurman also revealed that three other researchers from the foundation model team are about to leave.
The slow progress of Apple's AI has made team members lose confidence, leading them to leave one after another.
Apple's AI team has lost 10 members in just a few weeks, including its leader, Pang Ruoming.
Four Apple AI Researchers Leave
The news of Jian Zhang's departure has been confirmed by Meta, and he has also updated his LinkedIn profile accordingly.
At Apple, he served as the head of robotics research under the AIML team. This robotics research group is different from Apple's robotics product development department, which was merged into Apple's hardware engineering department earlier this year.
During his time at Apple, Jian Zhang's research focused on robot intelligence and human - robot interaction. He published several representative open - access papers and developed prototype systems, laying a complete technical foundation for Apple's robotics direction from perception - motion to emotional expression.
Jian Zhang graduated from Zhejiang University with a bachelor's degree and Purdue University with a doctorate. After obtaining his doctorate, he taught at Purdue for a year before joining Apple, where he has worked for ten years.
After joining Meta, Jian Zhang will serve as the chief director of robotics technology. His studio is part of Meta Reality Labs and is not part of the controversial MSL.
Besides Jian Zhang, John Peebles, Nan Du, and Zhao Meng from Apple's foundation model team are also reported to be leaving. However, they are not going to Meta. The first two will join OpenAI, and Zhao Meng will go to Anthropic.
John Peebles is one of the core members of the foundation model team. He is directly responsible for the training and inference infrastructure of large - language models and is also a co - author of the paper on the deep - learning training system AXLearn.
Interestingly, John Peebles and Bill Peebles from OpenAI's Sora team are brothers. Bill mentioned "brother John" in the acknowledgments of his paper.
Both brothers hold doctorates. John graduated from MIT, and Bill graduated from UC Berkeley.
They have taken different career paths from campus to work. However, at ECCV 2020, they co - authored the paper "The Hessian Penalty: A Weak Prior for Unsupervised Disentanglement" with researchers from Yale and Adobe.
Now, the two brothers will reunite at OpenAI.
Nan Du, who is also leaving Apple for OpenAI, has updated his LinkedIn profile to show his upcoming move to OpenAI.
Before leaving, Nan Du was a senior researcher at Apple's foundation model team, leading the design of KV - cache sharing for 3B edge models, 2 - bit quantization training, and parallelization of PT - MoE server models.
Nan Du graduated from the Georgia Institute of Technology with a doctorate. After graduation, he joined Google and then moved to Apple in October 2023.
Zhao Meng, who is going to Anthropic, is also a senior researcher at the foundation model team. His core responsibility was to train and optimize large - language models for Apple Intelligence.
Zhao Meng graduated from Peking University with a bachelor's degree and then obtained his master's and doctorate degrees from the Swiss Federal Institute of Technology in Zurich, graduating in 2024.
Before joining Apple, Zhao Meng worked as a software engineer at Google Research.
The Exodus from Apple Will Continue
Before these four researchers, six members of Apple's AI team had already switched to Meta, including Pang Ruoming, the head of the foundation model team.
In June 2024, Apple announced Apple Intelligence at WWDC2024, demonstrating a "smart assistant" scenario where Siri could access emails, messages, and maps across different apps.
In February 2025, Bloomberg reported that the redesigned Siri project had stalled due to engineering flaws and numerous bugs. The original early - testing plan was postponed, and the project was described as "likely to be delayed again."
To make matters worse, an investigation in April 2025 revealed that the core functions demonstrated at WWDC2024 had never been fully implemented internally. Team members only saw the demonstration script after the event. This "paper demonstration" severely undermined the trust of both the public and employees in the project.
On June 12, Apple postponed the "Siri AI upgrade" target to spring 2026, officially admitting that the annual mass - production plan had failed, and Apple Intelligence had entered a long - term delay.
Exposures in mid - August of the dual - prototype tests codenamed "Linwood" and "Glenwood" showed that Apple is continuing to refine its self - developed large - language model while also evaluating the possibility of integrating third - party models from Anthropic or OpenAI. Its self - developed model strategy has been forced to consider alternative options.
The negative news and slow progress over the past year have demoralized Apple's AI team. Coupled with Zuckerberg's generous offers, team members are leaving one after another.
Gurman also reported that according to informed sources, the exodus of Apple's AI employees is expected to continue...
Reference Links:
[1]https://www.bloomberg.com/news/articles/2025-09-02/apple-s-lead-ai-researcher-for-robotics-heads-to-meta-as-part-of-latest-exits
[2]https://www.bloomberg.com/news/articles/2025-08-22/apple-explores-using-google-gemini-ai-to-power-revamped-siri
This article is from the WeChat official account "QbitAI", written by Keleixi, and is published by 36Kr with permission.