On the evening of June 25th, news broke that DeepSeek, one of China’s leading general-purpose large model developers, has officially kicked off a massive hiring campaign. The goal? To at least double the size of every single department within the company. Digging into the details, this isn’t just about filling a few gaps; it’s a comprehensive sweep covering seven major categories: algorithms, R&D, operations, product management, data engineering, and back-office functions. They’ve opened up 33 specific positions right now, with opportunities available in both Beijing and Hangzhou—and yes, internships are welcome too.
Looking closely at the job descriptions, it’s clear DeepSeek isn’t just looking to hire a handful of extra coders. They are building an entire talent pyramid, aiming to cover the whole value chain from foundational model research all the way to frontend commercialization. Specifically, they are hunting for 8 full-stack and algorithm experts, 5 roles focused on model data strategy, and 4 positions dedicated to building the core AI systems.
This time around, the spotlight is heavily on “Agent” applications. DeepSeek is actively recruiting for General Agent Data Product Managers (covering office, lifestyle, and search domains). The mission statement for these teams is genuinely inspiring: “Text extended our memory, printing presses expanded the spread of knowledge, and the internet connected our senses. We believe Agents will be the next extension of humanity itself—breaking down the walls of static knowledge through search, reaching the frontiers of the digital world with tools, and shifting the human-AI relationship from simple conversation to true trust. This is how we truly liberate human action and creativity, building a new civilization of ‘human-machine symbiosis.’” To support this vision, they’re also looking for AI Product Operations specialists (focused on experience and service) and domain-specific data product managers for niche fields like minority languages, medicine, and law.
Beyond this official announcement, there’s been buzz since June 21st when Cui Tianyi, Head of the DeepSeek Harness team, posted a recruiting flyer for Agent roles that caught everyone’s eye. Cui was blunt about the situation: “As a newly formed department, DeepSeek Harness has big dreams but heavy workloads, and we desperately need people. I’m interviewing every single day and putting up flyers everywhere…” He outlined three key openings: Harness Researcher, Harness Engineer, and Harness Product Manager. According to him, while headcount needs are growing, the interview process remains standard—one written test followed by three rounds of interviews, with Cui himself handling the final round. All roles are based in Beijing.
Meanwhile, investors reacting to DeepSeek’s recent ~50 billion RMB financing round suggest that securing top-tier talent and computing power were likely the primary drivers behind the deal.
Back on June 16th, reports surfaced that DeepSeek’s parent company, Hangzhou Deep Sourcing (DeepSeek), completed its first external funding round. The total raised exceeded 50 billion RMB (roughly $7.4 billion), pushing the post-money valuation past $50 billion (about 338 billion RMB). Founder Liang Wenfeng reportedly invested about 20 billion RMB himself, making him the largest single investor. Tencent chipped in around 10 billion, while the CATL ecosystem (including CATL and Puquan Capital) contributed 5 billion. Other major backers included NetEase, JD.com, Monolith Lisi Capital, IDG Capital (each putting in roughly 3 billion), and Zhengxin Valley Investment and Shixiang Tech (each contributing about 1.5 billion).
However, growth brings challenges. From late 2025 to early 2026, at least five core R&D members have left DeepSeek to join other tech giants or top AI firms. For instance, Guo Daya, a core researcher for DeepSeek-R1, joined ByteDance Seed; Wang Bingxuan, the author of the first-generation large language model, moved to Tencent Hunyuan; Luo Fuli, a key developer of DeepSeek-V2, joined Xiaomi MiMo; Wei Haoran, who worked on the OCR series models, headed to Baidu; and Ruan Chong, a key contributor to multimodal results, started at the autonomous driving startup Yuanrong Qixing.
This exodus highlights a broader issue: a gap between talent supply and the exploding demand in China’s AI sector. According to a new economic talent report released by Maimai, from January to April 2026, AI-related job postings skyrocketed by 8.7 times, jumping from about 2.78% of all new economy jobs to a whopping 22.03%. AI has become the battleground for spring recruitment. Hiring remains heavily concentrated in first-tier and emerging tech cities, with Beijing, Shanghai, Shenzhen, and Hangzhou topping the list. Beijing leads the pack with an AI penetration rate of 30.17% in new job postings—meaning for every 10 new roles created, 3 are in AI.
In terms of companies posting the most new jobs, ByteDance takes the lead. DJI has jumped ahead of Xiaohongshu to take second place, while Xiaohongshu sits third. Giants like Tencent, Meituan, and Ant Group also feature prominently. Notably, AI-native players like MiniMax and Dreame have also cracked the top 20 for active hiring.
Despite the scramble for talent, founder Liang Wenfeng has made it clear at investor meetings that his commitment remains unwavering: continue developing open-source AI models while chasing the distant, ambitious goal of achieving Artificial General Intelligence (AGI). Liang has explicitly stated that his main focus is technological advancement, not immediate monetization. Given this aggressive roadmap, it makes perfect sense why the usually low-key DeepSeek feels compelled to accelerate its hiring engine.
The pace of innovation hasn’t slowed either. On April 24th, DeepSeek officially launched and fully open-sourced its V4 series of large models. These products utilize a 1.6 trillion parameter MoE (Mixture of Experts) architecture, natively supporting millions of tokens in context windows, with benchmark scores rivaling top foreign closed-source models. Just days later, on May 22nd, DeepSeek dropped a permanent API price cut announcement. The V4-Pro line saw billing standards slashed to just 25% of their original price. Cached input prices plummeted from 1 yuan per million tokens to a mere 0.025 yuan. Uncached input and output prices were simultaneously reduced from 12 yuan and 24 yuan to 3 yuan and 6 yuan respectively.
