He trained at Carnegie Mellon, then built one of China’s biggest AI wins
Moonshot AI co-founder Yang Zhilin's Carnegie Mellon-trained path to building Kimi K3 in China has sparked online debate about the US losing top AI talent.
Yang Zhilin, co-founder of Chinese AI startup Moonshot AI, is at the center of an online debate this week after his company released Kimi K3, a model it says performs nearly as well as Claude Fable 5 and GPT 5.6 Sol.
Zhilin completed his PhD at Carnegie Mellon University in just four years, jointly advised by Russ Salakhutdinov and William Cohen. During that time, he also worked at Google Brain and Meta Platforms, gaining experience inside two of the biggest US AI research operations.
Rather than staying in the US for postdoctoral work reportedly available at Stanford University or MIT, or joining Apple, which had tried to recruit him, Zhilin returned to China to build Moonshot AI.
The story resurfaced after Salakhutdinov posted on X congratulating his former student, writing that Zhilin ‘made truly fundamental contributions to ML’ during his time at CMU and calling his career ‘spectacular.’ The post prompted questions from social media users about why the US did not retain him.
Indian-origin entrepreneur Ankit Gupta was among those who responded, criticizing America’s visa policy and arguing the system fails founders. ‘The F-1 visa system is built around employment, not entrepreneurship,’ another commenter wrote, adding that it is ‘broken for founders.’
The debate comes as US authorities tighten F-1 visa rules, now capping stays at four years before an extension must be sought through the USCIS.
Wikimedia Commons/by Dllu
Leave a Reply