The person who helped build Baidu's flagship chatbot now carries a new title: vice president at a company that makes excavators.
Wu Tian climbed Baidu's AI hierarchy for years, eventually running its basic-model research unit and reporting directly to the founder.
Earlier this month she resurfaced at Sany Group, in a role that sounds less like a laboratory and more like a factory floor.
On 3 November, Wu announced on her social-media account that she had joined Sany Group (三一集团) as a group vice president, dean of the Yun AI Institute (耘AI总院) and, concurrently, dean of the company's Artificial Intelligence Research Institute. Baidu had confirmed on 29 October that she had left after close to twenty years. The move is one of the clearest examples yet of China's senior AI talent leaving internet platforms for industrial manufacturers.
Key takeaways
- Wu Tian (吴甜) announced on 3 November that she had joined Sany Group as a group vice president and dean of the Yun AI Institute (耘AI总院), while also heading the Artificial Intelligence Research Institute.
- Her former posts were group vice president at Baidu, head of its Basic Model R&D Department (基础模型研发部) and deputy dean of Baidu Research Institute (百度研究院副院长).
- She joined Baidu in 2006 after a master's degree from Zhejiang University, led the PaddlePaddle (飞桨) deep-learning platform and was one of the early technical leads on Ernie Bot (文心一言).
- Baidu confirmed her departure on 29 October; Sun Tianxiang (孙天祥) had taken charge of the basic-model unit in July 2026.
- The appointment points to a broader flow of AI researchers from consumer internet companies into industrial firms that want to put models into machines.
From basic models to basic machines
Wu's career at Baidu tracks the company's own AI history. She joined in 2006 after graduating from Zhejiang University, moved into its newly formed natural-language-processing group a few years later, and by 2018 was running the AI technology platform, covering language processing, knowledge graphs and the PaddlePaddle (飞桨) framework. She became a group vice president in 2020 and was among the early technical leads for Ernie Bot (文心一言).
In November 2025, Baidu created a Basic Model R&D Department (基础模型研发部) and put her in charge, reporting to chief executive Robin Li (李彦宏). Eight months later, in July 2026, the department passed to Sun Tianxiang (孙天祥), a 1997-born computer-science PhD from Fudan University who had worked on the MOSS model. Wu moved to the role of deputy dean at Baidu Research Institute (百度研究院副院长), which she held until leaving in late October.
Her new employer is a different kind of company. Sany was founded in 1989 and is headquartered in Changsha. It is best known for construction machinery — concrete equipment and excavators — and has spent years pitching itself as a manufacturer in transition, adding electrified machines, remote-controlled bulldozers and autonomous road-rollers. Its Yun AI Institute is the unit charged with turning that ambition into working systems, focused on AI in construction machinery and robotics.
Current and former titles
It is worth keeping the roles apart, because the two organisations are doing very different things.
Her current roles, as announced: group vice president of Sany Group, dean of the Yun AI Institute (耘AI总院) and dean of the Artificial Intelligence Research Institute. The institute sits inside the group's intelligent-strategy push and is expected to work on applied AI for heavy equipment and industrial robots.
Her former roles at Baidu: group vice president, head of the Basic Model R&D Department (基础模型研发部) from late 2025 to mid-2026, and deputy dean of the Baidu Research Institute (百度研究院副院长) from mid-2026 until October. The basic-model department she once led is now under Sun Tianxiang.
That separation matters. A headline that simply calls her a "former Baidu AI chief" blurs a chain of distinct jobs held over different periods, and it would overstate what she ran at the end.
What Sany is buying
The appointment is easy to read as a talent story. It is also a signal about where industrial AI is heading.
Sany has publicly framed 2026 as its first year of "intelligentization," and the company has said it aims to become an AI-native robotics business. Its machinery already includes remotely operated excavators and unmanned roller fleets used in ports and mines. Those systems need models that can handle changing sites, awkward physical tasks and equipment from many vendors — a different problem from answering questions on a screen.
Wu has spoken in the past about the need to combine general model capability with industry-specific data and knowledge, then support deployment through tools and platforms. That is close to the problem Sany is trying to solve. What remains unknown is the scope of her new remit: specific projects, team size and research plans were not disclosed in the announcements.
Honest limitations
This article draws on reports by Ta Kung Pao, the Hong Kong Commercial Daily and Shenzhen News, all of which describe the same announcement. Wu's move was made public through her own social-media post and a Baidu confirmation to a Chinese technology outlet; I have not seen the original post, only press accounts of it.
No direct quotation from Wu is used here. Chinese reports paraphrase a short line about starting a new phase with a newcomer's mindset, but without a verifiable primary text I have chosen not to reproduce it as a quote.
The reports differ slightly on her exact Baidu timeline and on whether she is described as a group vice president or a research institute deputy dean at the point of departure. I have used the sequence that appears in most accounts and flagged the distinction between current and former titles above.
Nothing in the sources explains why she left, how the move was negotiated, or what she will build at Sany. This article does not speculate on those points. Chinese names and institutional titles are translated roughly; the originals are kept where the wording matters.
What readers can do now
- If you follow AI talent moves, separate roles by date. "Head of basic models" and "deputy dean of a research institute" are different jobs with different authority, even at the same company.
- If you work in industrial technology, watch where model researchers go. Placements in machinery, energy and logistics companies are a practical signal of which sectors are hiring AI capability rather than just buying it.
- If you are hiring, note the pattern. Manufacturers are competing for people who can connect a general model to physical equipment — a scarcer combination than either pure research or pure operations.
Wu's move is not a verdict on Baidu or a guarantee for Sany. It is a useful marker: the people who built China's consumer AI are now being asked to put it to work in steel and concrete.
