Skip to main content

Baidu Hires Top AI Talent, DeepSeek Researcher to Lead Multimodal

Baidu is doubling down on its artificial intelligence ambitions with a pair of strategic hires aimed at strengthening its ERNIE large model. The company's Foundation Model Research Department (BMU) has brought in a senior researcher from a North American frontier lab, along with a former DeepSeek researcher who will now lead the multimodal algorithm team.

Sun Tianxiang, head of BMU, announced the new additions internally on September 3rd. The first is a researcher identified only as Wu Qin (a pseudonym), who has spent the past two years working on multiple generations of large models at a leading North American lab. Wu Qin's specific background remains under wraps, but his role will focus on enhancing ERNIE's pre-training capabilities—a critical step in improving the model's foundational understanding and performance.

The second hire is Wei Haoran, a former researcher at DeepSeek, who has officially transferred to Baidu and now heads the multimodal algorithm team for ERNIE. Wei joined Baidu in early 2026 and has already made an impact. In June, his team released the UnlimitedOCR model, which achieved the top global score on the OmniDocBench benchmark—a test that evaluates end-to-end OCR performance. This achievement set a new record and underscores the team's technical prowess.

Wei's background is impressive. At DeepSeek, he led the development of versions V3.2 and V4, as well as the DeepSeek-OCR series of open-source models. His experience in building multimodal large models is extensive, making him a valuable asset for Baidu's efforts to advance ERNIE's capabilities in understanding and processing visual and textual information.

These moves come at a time when Baidu is actively seeking to position itself at the forefront of AI development. Sun Tianxiang, who is relatively young at 27, took over as head of BMU in July and has since joined both the Baidu Model Committee and the Technical Strategy Committee. With the core team now in place, he is driving a dual upgrade—both organizational and technical—to ensure ERNIE stays competitive on the global stage.

The recruitment of top talent from international labs and competitors is a clear signal of Baidu's commitment to pushing the boundaries of what its large models can achieve. By aligning more closely with international frontiers in model development, Baidu aims to enhance ERNIE's performance across a range of applications, from natural language understanding to complex multimodal tasks.

For industry observers, these hires highlight the intensifying competition in the AI sector, particularly in the race to develop more sophisticated and capable large models. Baidu's focus on strengthening its research team is a strategic move to not only keep pace but also lead in innovation.

As ERNIE continues to evolve, the contributions of researchers like Wu Qin and Wei Haoran will be closely watched. Their expertise could prove pivotal in helping Baidu achieve its AI goals and deliver more advanced solutions to users worldwide.

Key Points

  • Baidu's BMU has hired a senior researcher from a North American frontier lab to boost ERNIE's pre-training.
  • Former DeepSeek researcher Wei Haoran now leads the multimodal algorithm team for ERNIE.
  • Wei's team released UnlimitedOCR, which topped the OmniDocBench benchmark in June.
  • Sun Tianxiang, head of BMU, is driving organizational and technical upgrades to align with international AI standards.