Yu Dong, Deputy Director of Tencent AI Lab, Leaves the Company; "New-Old Handover" in Hunyuan Team Underway | Exclusive Report from Intelligence Emergence
Text by | Zhou Xinyu
Edited by | Su Jianxun
According to multiple independent sources, Yu Dong, the former deputy director of Tencent AI Lab, will leave Tencent recently due to personal career development reasons.
As of the time of publication, Tencent has not responded officially.
As one of the earliest talent reserves for Tencent's large AI models, Yu Dong is a veteran in Tencent's AI ecosystem. After joining Tencent in 2017, Yu Dong served as a Distinguished Scientist, the deputy director of Tencent AI Lab, and the Chief Scientist of Tencent YouTu Lab. He was mainly responsible for the R & D of technologies related to speech processing, natural language processing, and digital humans. Prior to that, he had been engaged in R & D at Microsoft for nearly 20 years.
△ Yu Dong. Image source: Internet
Yu Dong is an expert in the fields of speech processing and deep learning. He once served as the chair of the IEEE Speech and Language Processing Technical Committee, published over 300 papers, and obtained more than 100 patents. He is one of the leading researchers who first successfully applied deep - learning technology in the field of speech recognition.
During his tenure at Tencent, Yu Dong led the research team to publish hundreds of papers in multiple top academic conferences and journals. He also promoted the application of NLP, speech, and digital - human related technologies in Tencent's business.
Yu Dong also made significant contributions to the R & D of Tencent's large model, "Hunyuan". The "Hunyuan" team belongs to Tencent's Technology Engineering Group (TEG), spanning multiple departments such as Big Data, AI Lab, and Machine Learning Platform Department. In the R & D system of Hunyuan, Yu Dong was also responsible for multimodal generation and understanding, as well as some text - related research work.
In building the talent system for Hunyuan, Tencent doesn't dare to slack off. Even with the departure of a "veteran", there is still a continuous influx of fresh blood.
Since 2025, with the aggressive entry of DeepSeek, the major tech companies have quickly reached a consensus: Foundation models are the core competitiveness, and the capabilities of foundation models determine the upper limit of the user experience of AI applications. Recently, focusing on the R & D of large models, Intelligent Emergence reported exclusively that Tencent is making a series of internal adjustments:
On the one hand, Tencent is bringing in new talent and increasing investment in human resources. In the second half of 2025, after Yao Shunyu, a former researcher at OpenAI, joined Tencent and took up positions such as the Chief AI Scientist in the "CEO/President's Office", Hunyuan quickly attracted several core employees from companies like ByteDance, Alibaba, and Dark Side of the Moon.
On the other hand, Tencent has also integrated the R & D resources for large models internally.
Previously, the model R & D teams and resources were scattered across different business groups, mainly to meet the needs of different businesses. However, this also led to a lack of focus in R & D efforts and a waste of resources. In the case of TEG, the R & D of Hunyuan spanned multiple departments such as the Large Language Model Department, AI Lab, and Machine Learning Platform.
Recently, in order to integrate model R & D resources and clarify the division of labor in model R & D, Tencent established new departments in TEG, including the AI Infra Department, AI Data Department, and Data Computing Platform Department.
"The market is still in a highly competitive stage without an absolute leader," said Liu Chiping, the president of Tencent Group, at the Q3 2025 earnings conference call. The results of the adjusted model R & D will be crucial for Tencent.
Welcome to communicate!