DeepSeek's "Ingenious Chess Game"
```
DeepSeek, which prefers to keep a low profile, may be shifting towards a more assertive stance.
With the release and open-sourcing of its next-generation large language model V4, market attention has quickly focused on DeepSeek’s latest product.
What few people have noticed is that, on the same day V4 went online, DeepSeek also released a series of job openings, covering positions from technical R&D to management and finance.
From the release of financing information to the launch of its next-generation data model, establishing a data center in Inner Mongolia, to the introduction of various job openings, DeepSeek seems to be making more strategic moves on a much larger “chessboard” than before.
Registered capital quietly changes
In April, the hottest fundraising news in the industry was DeepSeek. Industry rumors say the company is in talks with investors to raise funds at a valuation of about $10 billion, with the deal size not less than $300 million.
Getting a “ticket” to enter China’s large language model market at this valuation would inevitably spark fierce competition in the industry.
It’s worth noting that since its founding, DeepSeek has never brought in external capital, with funding mainly relying on continuous investments from its controlling shareholder Liang Wenfeng and his affiliated institution, Qianfang Quant.
Interestingly, Tianyancha showed on April 27 that DeepSeek’s parent company, “Hangzhou DeepSeek Artificial Intelligence Fundamental Technology Research Co., Ltd.,” underwent an adjustment in registered capital.
The company’s registered capital increased from 10 million RMB to 15 million RMB. The founder, Liang Wenfeng’s subscribed capital rose from 100,000 RMB to 5.1 million RMB, and his direct ownership stake rose from 1% to 34%. Ningbo Cheng’en Enterprise Management Consulting Partnership (Limited Partnership) saw its holding drop from 99% to 66%.
By combining both direct and indirect holdings, Liang Wenfeng now controls about 84.29% of the parent company’s equity.
It is unknown whether this capital increase and the rise in founder shareholding is related to rumored considerations of introducing external capital.
Intensive recruitment of Agent roles on V4’s launch day
Zhishi Tang noted: On the day V4 launched, DeepSeek released three types of Agent-related positions: Deep Learning Algorithm Researcher, Full Stack Developer, and Data Strategy Engineer.
Agents play a central role in today’s large model competition: not only responsible for model generation, inference, and multitasking, but also for complex instruction understanding and application—they are a key driver for commercialization of the models.
The timing of these postings shows the company immediately strengthened this function after V4’s launch, reflecting the importance of Agent capabilities to the realization and competitiveness of the new version.
In terms of responsibilities, the Deep Learning Algorithm Researcher focuses on model inference, generation, and research into reinforcement learning methods, enhancing Agent core capabilities; the Full Stack Developer builds Agent platforms, manages containers and toolchains, optimizing development and deployment; the Data Strategy Engineer designs high-quality training and evaluation data end-to-end, ensuring Agents are stable and efficient in diverse scenarios.
The professionalism and division of labor in these roles hint at an internal strategic signal: DeepSeek is establishing a complete organizational loop to ensure the implementation of Agent capabilities.
This further indicates that DeepSeek attaches great importance to Agent capabilities, seeing them as a key support for commercialization.
First appearance of “single-location position” in Hangzhou?
After V4’s launch, DeepSeek released a previously unseen position that afternoon—AI Supercomputing Cluster Operations Engineer. The main responsibility is to ensure the stable operation of large model training and inference platforms, including daily inspections, troubleshooting, system performance optimization, and constructing automation tools to improve operational efficiency.
Compared to algorithm and frontend positions released in both “Beijing + Hangzhou,” this position appears only in Hangzhou, indicating the company is operating or intends to operate a local data center and needs a specialized team for large-scale computing power.
The “plus points” in the job description also reveal key signals: experience with AI supercomputing clusters above 1,000 cards, GPU server deployment, and performance optimization—meaning the company wants an industrial-grade, sustainably-operating high-performance computing environment, not just conventional server management.
Of note: describing AI supercomputing clusters above 1,000 cards implies DeepSeek has configured over a thousand high-performance GPUs to support V4’s launch.
This further suggests post-launch V4’s demand for computing power is extremely high, and ordinary technical staff can’t support it—more specialized talent is required.
From the overall layout, Beijing mainly handles algorithm R&D and system development, while Hangzhou is responsible for supercomputing cluster operations. The two-location setup forms a complementary closed loop. The supercomputing cluster is not just hardware but additionally represents DeepSeek’s strategic support for V4’s implementation and intensive computation, indicating internal organizational adjustments tailored for large model operations.
“Expansion” signals from the company
After V4's launch, in addition to technical and R&D positions, DeepSeek also intensively supplemented management and organizational roles.
Zhishi Tang found that this large model company has highly divided functional support positions into three directions: Recruitment Manager, Comprehensive Service HR, and Corporate Culture Supervisor.
There are also several “timing signal” details worth noting: the Recruitment Manager position was posted a week before V4 launched, responsible for devising high-potential talent recruitment strategies, tracking recruitment progress, and analyzing industry trends; Comprehensive Service HR and Corporate Culture Supervisor roles were posted the afternoon V4 launched. These respectively handle employee transfers, compensation, performance, training, and major event planning, as well as strengthen cross-department collaboration in daily operations.
These developments show that after launching the new version, the company not only faces technological iteration challenges but also new pressures on internal management.
As R&D and product team sizes rapidly expand, staff numbers continue to grow, and individual positions struggle to cover recruitment, daily HR, and cultural responsibilities. This suggests DeepSeek is entering a “complex phase” of rapid expansion.
Coincidentally, DeepSeek also posted an Accountant position, mainly responsible for daily accounting, tax filings, asset management, and cooperation with external tax agencies. This setup shows the company is improving its financial system to ensure cash flow, cost control, and tax compliance.
Risk Reminder and DisclaimerThe market carries risks; investment requires caution. This article does not constitute personal investment advice and does not take into account the specific investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article fit their specific circumstances. Invest at your own risk. ```