Alibaba Loses AI Talent as Rivals Raid Tongyi Lab
In recent weeks Alibaba’s high-profile Tongyi Lab — the team behind the Qwen family of open models — has seen several senior researchers depart for competing companies. The wave of exits, which includes veterans who helped build the lab’s voice and vision capabilities, has sparked fresh headlines about an escalating talent war among China’s biggest tech players.
The departures are notable both for who is leaving and where they are going. Industry reporting shows that some former Tongyi leaders have moved to JD.com’s research arm, while others have joined Tencent’s growing Hunyuan large-model effort. These moves underscore how aggressively rivals are recruiting experienced engineers and researchers who can accelerate product roadmaps and strengthen proprietary stacks.
Names and Moves
One of the most visible departures is Yan Zhijie, a senior figure who led voice efforts at Tongyi and is reported to have joined JD’s Explore/Discovery research institute as head of a voice lab. Another prominent exit is Bo Liefeng, who had responsibility for applied vision work at Tongyi and is now linked with Tencent’s Hunyuan team. These transitions are already prompting internal reorganizations and new hires at Alibaba to fill gaps in critical areas such as speech, multimodal perception and production-grade system engineering.
Why Talent Is Leaving
Multiple forces are driving the churn. First, the market for top AI talent is extraordinarily tight: engineers and researchers who have experience building and deploying large models are scarce, and companies are willing to offer generous packages and leadership roles to attract them. Second, corporate reorganizations and shifting strategic priorities — including changes inside research groups and lab leadership — make high-quality staff more open to new opportunities. Finally, the open-source success and wide adoption of models like Qwen increase the visibility and market value of researchers who contributed to them.
How Alibaba Is Responding
Alibaba has not stood still. The group has been filling roles and recruiting experienced specialists to lead voice and perception teams — including bringing in engineers with backgrounds at Baidu, Didi and promising startups — to reconstitute capacity quickly. The company’s cloud and research arms continue to iterate on Tongyi Qianwen and related products, even as they adjust staffing and project scopes to retain momentum.
What This Means for the Chinese AI Ecosystem
The high-velocity hiring cycle has broader implications. When leading researchers move between giants such as Alibaba, Tencent, JD and ByteDance, knowledge and best practices diffuse rapidly across the ecosystem, accelerating product development but also heightening strategic competition. Companies are also expanding campus recruitment, internships and training programs to build long pipelines of junior talent who can eventually shoulder advanced research work.
At the same time, established firms are investing in alternative approaches to talent scarcity: partnerships with startups, acquisitions, and building industrial research alliances that can pool expertise. This “both/and” strategy — buy where needed, train where possible — reduces immediate exposure while laying the groundwork for longer-term independence.
Strategic Takeaways
For Alibaba, the exits are a reminder that leadership in models and applications requires not only strong algorithms but also continuity of teams that understand production constraints, data pipelines and platform integrations. For rivals, hiring seasoned Tongyi alumni offers a fast track to shore up their product stacks and credibility in key domains like speech, vision and multimodal systems.
The current talent scramble will likely continue so long as demand for large-model expertise outpaces supply. That makes investments in retention, culture, and career pathways as strategically important as raw compensation — and it will shape where China’s next wave of AI products emerges.