본문 바로가기
bar_progress

Text Size

Close

Identity of 'DeepSeek' Founder Behind Nvidia Plunge... Born in 1985, Graduate of Prestigious Chinese University

Liang Yuanfeng, Computer Engineering Major at Zhejiang University
Founded a Hedge Fund with Two College Friends and Achieved Remarkable Success

Attention is focused on the founder of the Chinese AI startup 'DeepSeek,' which has caused a stir in the global artificial intelligence (AI) industry. The founder, Liang Yuanfeng, born in 1985, is known as a finance professional who majored in computer engineering.


The US daily newspaper The Wall Street Journal (WSJ) reported on the 27th (local time) that DeepSeek was established in May 2023 in Hangzhou, China. Founder Liang Yuanfeng is from Guangdong Province and majored in computer engineering at the prestigious Zhejiang University in China.


Identity of 'DeepSeek' Founder Behind Nvidia Plunge... Born in 1985, Graduate of Prestigious Chinese University AI model of Chinese AI startup DeepSeek. APF=Yonhap News

After graduating from university, he founded a hedge fund called 'High-flyer' with two friends. The hedge fund primarily used so-called 'algorithmic trading,' an automated trading method utilizing advanced computer equipment. In particular, Liang Yuanfeng introduced cutting-edge AI models such as deep learning into High-flyer's algorithms, which helped the assets under management (AUM) grow to about 8 billion dollars (approximately 11.5 trillion won).


Following High-flyer's success, Liang Yuanfeng created and operated a small AI research institute, which later spun off as an independent corporation; this company is the DeepSeek Research Institute that released the DeepSeek model.


Since 2019, High-flyer has been stockpiling NVIDIA graphics processing units (GPUs) for AI development. It is said that the company currently possesses an AI chip cluster equipped with about 10,000 GPUs.


DeepSeek released its first open-source AI model, 'DeepSeek Coder,' in November 2023, and in May last year, it unveiled the more advanced 'DeepSeek-V2.' The model attracted public attention as it was trained using only 2,048 H800 GPUs, which are known to perform significantly worse than the expensive H100 GPUs used by most big tech companies. DeepSeek claims that the model training cost was only 5.5 million dollars (approximately 8 billion won).


Jack Carse, a former executive of the leading US AI company OpenAI, emphasized about DeepSeek that "resource constraints often foster creativity, which is a great lesson."


© The Asia Business Daily(www.asiae.co.kr). All rights reserved.

Special Coverage


Join us on social!

Top