In an exclusive development, it has been learned that Qiyin Wu, a former Staff Engineer at Google DeepMind and a core contributor to the Gemini large model project, has recently joined Baidu to lead pre-training initiatives for large-scale models. Her previous rank at Google is estimated to be L6-L7, which corresponds to Alibaba's P9/P10 or ByteDance's 3-2/4-1, and she will be based in the United States for her new role.

Qiyin Wu holds a Master's degree in Computer Science from Stanford University. Before advancing to Staff Engineer at Google DeepMind, she accumulated extensive experience across various core technical teams at Google and contributed to engineering projects at the University of California, Berkeley (UC Berkeley). Notably, this Chinese female scientist was a lead author and key co-contributor to the official technical report for the Gemini 2.5 series models released by Google DeepMind.

Just two days prior, a report highlighted that another Gemini pre-training specialist, ranked L7-L8 at Google, had recently joined a major Chinese internet company, based long-term in the UK. This raises the question: why are Chinese tech giants simultaneously targeting talent from Google?

On one hand, top-tier overseas pre-training talent can help major domestic companies navigate the challenges of approaching the scaling law ceiling with fewer detours. On the other hand, ongoing organizational instability at Google—marked by the departure of Chief Scientist Jeff Dean from Google DeepMind—has gradually diluted DeepMind's status as the premier overseas laboratory, inadvertently opening a window for talent movement.

The race in large model development is a contest of computational power, but even more so, of talent. The significant premiums generated by this talent migration ultimately reflect a substantial trade-off in computing costs. Whether Baidu's intensified investment in AI talent for large models will translate into technical momentum remains to be seen over time.