Chui Tianyi, head of the DeepSeek Harness team, publicly recruited on social media, releasing about 150 positions at once, stating that "many new directions, new systems, and new demands need to be addressed."

For DeepSeek, which is known for its "small team and high talent density," this is not a small recruitment. It is also another public expansion following the large-scale recruitment signal released in June this year.

More Focused: Hiring Only for Backend Development and Agent Elastic Computing

The focus of the two rounds of recruitment differs. In the previous round, DeepSeek aimed to "expand all departments by at least double," covering seven categories of positions. This time, it is more focused, mainly hiring backend development engineers and agent elastic computing research engineers.

Chui Tianyi stated that this is a long-term large-scale investment. Senior engineers with 2 to 10 years of experience are the main target, and fresh graduates and those with less than two years of experience are also welcome. The age requirements for particularly outstanding candidates can be appropriately relaxed. The job locations are mainly Beijing, with some options available in Hangzhou. He called this recruitment "an unprecedented hiring demand for 150 HC."

Scale expansion is the main reason. Chui Tianyi explained that data volume, machine and container numbers, training tasks, evaluation tasks, agent environments, user and request volumes are all growing rapidly. "Any aspect in computer science will bring a significant increase in complexity when the scale increases," and the existing backend system may not be able to continue supporting future needs. Therefore, a large number of people need to be hired to upgrade, maintain, and rewrite various backend systems.

Expansion Continues, Model Rankings Have Dropped Slightly

Over the past three months, DeepSeek has been active: after completing a $5 billion funding round in June, it first launched a large-scale recruitment, then continuously updated the official versions of V4-Flash, V4 Pro, and the developer preview version of Harness, while significantly increasing API prices, with the highest increase reaching 11 times. Recently, there were reports that DeepSeek plans to deploy at least 160,000 Huawei Ascend 950DT chips in its newly built data center in Wulanchabu, Inner Mongolia, which could become one of the largest domestic AI chip clusters known so far, but both parties have not officially responded. In terms of capital, market sources say that DeepSeek has started its second round of financing, planning to raise 5 billion yuan, with a pre-money valuation of around 500 billion yuan, and its ARR is reportedly as high as 500 million US dollars.

On the user side, according to Quest Mobile data, as of June 2026, DeepSeek's monthly active users reached 129 million, ranking third domestically, behind Doubao's 382 million and Qianwen's 167 million. However, according to Artificial Analysis's intelligence index, DeepSeek scored 36 points and ranked 15th, with Zhipu and Kimi ranking 7th and 9th among domestic models, respectively.