DeepSeek, a Chinese AI startup founded by Liang Wenfeng, is making waves in the tech world with its groundbreaking V3 large language model (LLM). The company, based in Hangzhou, has managed to develop a powerful AI model despite operating with limited resources. Their latest achievement is the release of DeepSeek V3, which, trained on fewer resources, has performed as well as, or even exceeded, the AI models of major US competitors like Meta Platforms and OpenAI.

The success of DeepSeek is largely attributed to its innovative approach to talent. The company’s preference for hiring fresh graduates or individuals early in their AI careers has become a defining characteristic. In fact, most of the developers at DeepSeek have only one or two years of work experience. According to Liang Wenfeng, the founder, this approach prioritizes talent and ability over experience, which contrasts with the typical hiring practices of other AI startups in China.

DeepSeek’s team includes young innovators like Gao Huazuo, a graduate of Peking University, and Zeng Wangding, a master's student from Beijing University of Posts and Telecommunications. These individuals have played a crucial role in the development of DeepSeek’s MLA architecture. This talent-driven strategy has allowed DeepSeek to create an AI model that rivals the most advanced systems despite limited resources.

The DeepSeek V3 model’s development was a monumental task completed in just two months, using only $6 million and approximately 2,000 Nvidia H800 chips—far fewer than those used by its competitors. Despite this, the model's performance has stunned experts in the field. Andrej Karpathy, a founding team member at OpenAI, described the budget as a "joke," yet the model achieved impressive results, proving the effectiveness of DeepSeek’s approach.

Another standout figure in the company’s rise is Luo Fuli, a former employee known as an "AI prodigy" by Chinese media. Luo, who was offered a significant annual salary by Xiaomi founder Lei Jun, has helped raise DeepSeek’s profile as an AI powerhouse.

Leading this team of young talent is Liang Wenfeng, whose leadership style has been described as more mentoring than authoritative. Former employees note that Liang often proposes solutions in a suggestive manner, which has allowed the team to innovate freely and contribute to the company’s success.

SOURCE: SCMP | PHOTO: LATENODE