People move, and this may be the person Lei Jun values most right now. Xiaomi recently circulated an internal promotion list, with 31-year-old large model team head Luo Fuli elevated to Level 22. That means from her official onboarding in November 2025 to reaching the ceiling of Xiaomi's job hierarchy, she needed only 10 months. In doing so, she set Xiaomi's fastest promotion record in history. A woman from Yibin, Sichuan, she earned her bachelor's degree from Beijing Normal University and her master's degree from Peking University. Her best-known experience was at DeepSeek, where outsiders branded her a "prodigy girl," and she was ultimately recruited by Lei Jun with a hefty offer. Yet she has said more than once that she is not a prodigy and simply wants to do her work quietly. Viewed more broadly, Luo Fuli's exceptional promotion is a true microcosm of the talent competition in China's AI industry. These young people are leading the direction of the tide in China's AI sector.

Lei Jun Poached a Post-95s Woman, and Her Rise Is the Fastest. In just 10 months, Luo Fuli's Xiaomi career has been unstoppable. Alongside the promotion news came the MiMo team's official release and open-sourcing of the Xiaomi MiMo-V2.6 series, including two native omni-modal models, Pro and Flash. Thanks to expanded RL compute, MiMo-V2.6-Pro scored 46 on the Artificial Analysis Intelligence Index, ranking first among open-weight models; at comparable intelligence levels, its price is only 1/20 to 1/60 that of overseas models. The MiMo team sees this as a key step in exploring the RSI (recursive self-improvement) path. Luo Fuli admitted, "The research innovation and engineering challenges behind this work exceeded those of DeepSeek R1, in which I was partially involved." According to public information, apart from Xiaomi founder Lei Jun, Xiaomi Group has ten job levels from 13 to 22, with directors corresponding to Levels 19 to 20 and vice presidents to Levels 21 to 22. A Level 19 employee's total annual compensation can exceed RMB 2 million, which means Level 22 pay must be even higher. In other words, after leading her team through the release of the new model, Luo Fuli has joined the ranks of Xiaomi's core senior executives.

Tracing how Luo Fuli joined Xiaomi, it starts with a trending topic. Early last year, according to Securities Times, Lei Jun had hoped to poach Luo Fuli with an annual salary of tens of millions of yuan, which unexpectedly made her famous. Then in November last year, she posted on WeChat Moments that "intelligence will move from language to the physical world," announcing she was joining Xiaomi's MiMo team. Soon after, Luo Fuli completed her first public appearance at Xiaomi. At Xiaomi's "Human x Car x Home" ecosystem partner conference in December last year, she made her formal debut as head of the MiMo large model. In March this year, Xiaomi released three models at once, including the flagship foundation model for the agent era, Xiaomi MiMo-V2-Pro, as well as the omni-modal foundation model Xiaomi MiMo-V2-Omni and the speech synthesis model Xiaomi MiMo-V2-TTS. At the time, Luo Fuli posted on social media, "I call this a silent ambush — not because we planned it, but because the shift from Chat to the Agent paradigm happened so fast that even we could hardly believe it. Somewhere in the middle, there was a process that was exciting, painful, and fascinating." In July this year, Xiaomi adjusted the organizational structure of Xiao Ai, splitting technical capabilities into three parts: foundation models, cloud, and on-device. Among them, foundation model capabilities are provided by the MiMo team led by Luo Fuli, which also shows that her business scope has further expanded. Up to this promotion, Luo Fuli used just 10 months to reach the ceiling of Xiaomi's job hierarchy. Lei Jun's hunger for young AI talent is self-evident. Promotion: Holiday PCE and nonfarm payrolls take turns on stage! Gold faces a major test. View full content. Liang Wenfeng could not keep her: Luo Fuli's growth story.

Behind Luo Fuli is not the kind of legendary story of early fame that outsiders imagine. Born in 1995 into an ordinary family in Yibin, Luo Fuli attended Yibin No. 1 Middle School for high school. Her homeroom teacher once commented that she was "not top-notch, but hardworking enough." Later she entered Beijing Normal University to study computer science, which in her own words made her "an absolute 'low-starting-point person' in the computer field." Even so, Luo Fuli began preparing in her sophomore year to pursue a recommended graduate spot in computer science at Peking University, and through hard work she eventually got in. At first, her research interest focused on a subfield of natural language processing, word sense disambiguation. But with the rise of large model technology, she judged that the essence of traditional disambiguation tasks was being covered by the large model paradigm of "next-word prediction." So she decisively shifted her research focus to text generation, which also laid the groundwork for her later work on large models. The year Luo Fuli truly emerged was 2019, when she published 8 papers at ACL, a top international conference in natural language processing, 2 of them as first author. The "prodigy girl" label followed. But Luo Fuli had an almost instinctive rejection of that label. She said bluntly in an interview with the Peking University alumni association that such labels are stereotypes created to attract attention, "whitewashing the real effort, resilience, and full commitment in the research process, implying that success is innate." Recalling her years as a graduate student at Peking University, Luo Fuli once said, "Peking University's free research environment benefited me greatly." At the same time, Peking University also gave her her earliest "collaboration enlightenment." In the lab there, she first experienced the power of teamwork: some classmates were good at building GPU clusters, some specialized in algorithms, and everyone displayed their strengths. After finishing her master's degree, Luo Fuli chose Alibaba as the first stop in her career. She joined Alibaba Damo Academy through the "Alibaba Star" program and served as a researcher in the Machine Intelligence Lab, conducting research in artificial intelligence. There, she led the development of the multilingual pre-trained model VECO and promoted the open-sourcing of AliceMind. Three years later, Luo Fuli joined DeepSeek's parent company High-Flyer Quant to work on deep learning-related matters, and later served as a deep learning researcher at DeepSeek, participating in the development of models such as DeepSeek-V2. With the sudden emergence of the DeepSeek large model, Luo Fuli began to appear more often in the spotlight. She once summarized, "That period at DeepSeek was a major turning point." Looking back, there was no shortcut or luck in Luo Fuli's growth path. It is a typical result of long-termism and the compounding of technical skills, and she has always maintained a self-awareness that de-mythologizes herself. As she said, she is not a prodigy and her starting point was actually quite ordinary: "It was curiosity in exploring essential questions, decent execution, and a bit of unwillingness to lose that brought me step by step to where I am today."

The AI Battle and Elite Talent: A Golden Age. A war for AI talent is quietly emerging. It is reminiscent of an earlier scene: last December, Tencent appointed former OpenAI researcher Yao Shunyu as chief AI scientist, when he was only 27 years old; afterward, the Hunyuan multimodal model department and the large language model department merged to form the Foundation Model Department, managed uniformly by this young man. ByteDance has also gone all in. Early this year, DeepSeek core researcher Guo Daya resigned, triggering a chase by a host of internet giants, and he ultimately chose to join ByteDance's Seed department. At the same time, ByteDance launched the "Top Seed" campus recruitment program, recruiting 30 top fresh PhDs globally and emphasizing industry-leading compensation with no upper limit. Baidu also joined the contest. In July this year, Sun Tianxiang, born in 1997, formally joined Baidu as head of foundation model R&D and also entered Baidu's model committee. A computer science PhD from Fudan University, he was the first core developer of the ChatGPT-like open-source large model MOSS and was the first internationally to propose the Model-as-a-Service concept. Before young people even leave campus, big tech companies are already scrambling to extend olive branches. Tencent officially launched the "Qingyun Plan" internship recruitment in March this year, opening six major technical fields including large language models, multimodality, and agents, with a compensation policy that has no cap; Moonshot AI announced the Kimi top talent "Time Travel Plan," granting company options to interns who have not yet graduated; MiniMax also launched the Top Talent program, and CEO Yan Junjie publicly stated that for truly top talent, salary and options have no upper limit, seniority is not considered, and only how big a problem you can solve matters. As early as the end of 2025, Business Insider disclosed a set of data: a 4-6 month AI internship and short-term research project had monthly pay in the USD 7,000-18,000 range, equivalent to about RMB 49,000-126,000. According to LatePost, ByteDance Top Seed campus hires had annual salaries of about RMB 1.5 million in 2024, rising to RMB 3-5 million in 2025, and some core positions in 2026 were offered as much as RMB 6 million, doubling year after year. There is no doubt that the domestic competition for AI talent has been fully upgraded, from early high-salary poaching and welfare competition to a comprehensive talent-heavy model of strategic delegation, resource tilting, and exceptional job levels. This also marks that young technical talent has become the core competitiveness determining AI companies and influencing the direction of the industry landscape. The paths of people such as Luo Fuli and Yao Shunyu offer young people a reference that is precisely not a "genius narrative," but a more plain logic: continuously accumulate engineering capability in the right direction, make frontier technology deployable, and then amplify your value on a sufficiently large platform. As Luo Fuli advised young people at the BAAI conference: always maintain the desire to explore and curiosity. "Use frontier AI tools to the utmost, and in the process of continuous trial and error, cultivate your own judgment and research taste. In an era of rapid technological change, unique cognition, judgment, and taste are the most core and irreplaceable competitive advantages for young people." In the final analysis, AI competition has never been only a contest of models and compute, but a contest of talent density. It can be asserted that the next change in the AI landscape will not happen at a single company's launch event, but at those moments when these young people decide "what to do and what not to do." The direction they choose is the direction of China's AI.