Alibaba recruits top Chinese AI scientist Li Xiangang to lead speech-recognition push

South China Morning Post

18-06-2025

Li Xiangang, a leading Chinese scientist in speech recognition, has joined Alibaba Group Holding to spearhead its artificial intelligence (AI) voice team, boosting the tech giant's capabilities in the burgeoning field.
Sources familiar with the matter said Li, who holds a PhD in Computer and Information Sciences from Peking University, had taken on a role leading speech AI research at the Hangzhou-based e-commerce giant. He fills a position previously held by Yan Zhijie, who left the company.
Alibaba owns the South China Morning Post.
The speech team, part of Alibaba's Tongyi Lab, focuses on multimodal speech and language models. In July 2024, the lab open-sourced two foundational speech models, SenseVoice and CosyVoice. SenseVoice's multilingual speech recognition notably outperformed OpenAI's Whisper by 50 per cent in Chinese and Cantonese, according to Alibaba.
Li's move was first reported by Chinese media outlets.
Machine-learning models for speech and recognition have broad AI applications, including in chatbots and digital avatars. Photo: Shutterstock Images
Machine-learning models for speech and recognition have broad AI applications, including in chatbots and digital avatars. This has led to intense competition among China's Big Tech firms wanting to stake out their positions in the sector. Chinese search giant Baidu, for example, showcased a digital avatar of Luo Yonghao, a prominent Chinese public speaker and entrepreneur, during its AI Day on Tuesday.