Liang Wenfeng

Some Lesser Known Facts About Liang Wenfeng

• Liang Wenfeng grew up in Guangdong, China.

• While many people around him focused on starting businesses, he was more focused on academics from a young age.

• His master’s thesis focused on developing a target-tracking algorithm using an affordable PTZ camera. His work, guided by Professor Xiang Zhiyu, explored ways to improve tracking accuracy with low-cost equipment.

• In 2008, Liang led a team of classmates to gather financial market data and explore quantitative trading with machine learning, during the peak of the 2007-2008 financial crisis.

• After graduation, Liang moved from Hangzhou to Chengdu, Sichuan, where he settled in a low-cost apartment. He later began exploring ways to apply AI in different areas. However, his experiments did not work out until he shifted his focus to finance.

• In 2013, Liang partnered with Xu Jin, a fellow Zhejiang University graduate, to combine AI with quantitative trading, which led to the launch of Hangzhou Yakebi Investment Management Co. Ltd.

• In 2015, Liang and Jin co-founded Hangzhou Huanfang Technology Co. Ltd., which was later renamed as ‘Zhejiang Jiuzhang Asset Management Co. Ltd.’

• In February 2016, Liang and two of his engineering classmates co-founded Ningbo High-Flyer Quantitative Investment Management Partnership. They used mathematics and AI as the foundation for their investment strategies.

• In 2019, Liang established a hedge fund and artificial intelligence company named High-Flyer to develop AI algorithms and explore their fundamental uses.

• Over time, High-Flyer became the first quantitative hedge fund in China to raise over 100 billion yuan ($13 million).

• When Liang first approached his partners at High-Flyer, they were not impressed. He seemed like a nerd with a strange hairstyle who could not explain his vision clearly. Liang said he wanted to create something groundbreaking, but his partners thought only big companies like ByteDance or Alibaba could achieve it.

• On 30 August 2019, Liang spoke at the Golden Bull Awards about ‘The Future of Quantitative Investment in China from a Programmer’s Perspective’. He explained that in quantitative investing, decisions are made by algorithms, not people.

  

• Liang’s speech at the Golden Bull Awards, held in Malaysia, led to a lot of discussion. During his speech, he also mentioned High-Flyer’s goal to improve efficiency in China’s secondary market.

• In February 2021, Liang wrote the preface for the Chinese edition of Gregory Zuckerman’s book ‘The Man Who Solved the Market: How Jim Simons Launched the Quant Revolution.’ In it, Liang shared how Jim Simons’ quote “There must be a way to model prices” inspired him whenever he faced challenges at work.

• In 2021, while running High-Flyer, Liang started buying thousands of Nvidia GPUs (powerful chips used for AI and data processing) for his side project. Some in the industry saw this as a billionaire’s quirky hobby.

• In May 2023, Liang announced the creation of DeepSeek to develop artificial general intelligence and maximise the use of available processors in China.

• In May 2023, Liang revealed that High-Flyer had purchased 10,000 Nvidia A100 GPUs before the US imposed chip restrictions on China, which limited China’s access to advanced AI chips. This early purchase gave DeepSeek a strong foundation for developing as an LLM company.

• Since investors were hesitant to invest in a long-term project, DeepSeek relied on funding from High-Flyer. It prioritised hiring talented and passionate employees over those with just experience.

• In November 2023, DeepSeek made a big impact in the industry by launching its first AI model, DeepSeek Coder.

• It then released two more models: the powerful DeepSeek LLM with 67 billion parameters, and the affordable DeepSeek-V2. The low prices of these models sparked a price war in China’s AI market, causing big companies like ByteDance, Tencent, and Baidu to lower their prices too.

• In a July 2024 interview with 36Kr, Liang said that China’s real challenge in innovation is organising talent, not funding. He also mentioned how DeepSeek hires local talent and believes keeping technology closed-source only slows competition briefly.

• On 20 January 2025, Liang attended a symposium in Beijing hosted by Premier Li Qiang. He was invited as an expert to share his views on the 2024 government work report.

  

• In January 2025, DeepSeek introduced R1, an open-source AI model with 671 billion parameters. They also shared a technical paper explaining how it was built.

• DeepSeek’s R1 model was developed using 2,048 Nvidia H800 GPUs, costing $5.6 million, offering a more affordable alternative compared to the expensive models from Western companies.

• By 27 January 2025, DeepSeek became the top free app on the U.S. iOS App Store, surpassing ChatGPT. This led to a huge drop in U.S. stocks with over $1 trillion lost in market value due to concerns over DeepSeek.

• DeepSeek’s growth within just one week of launching caught the attention of US President Donald Trump, who referred to it as a “wake-up call” for American tech companies.