Reported by NUPIAO
Just recently, a recruitment poster for an Agent position dropped by Cui Tianyi, the head of the DeepSeek Harness team, stirred up quite the buzz. The job requirements explicitly listed “good Chinese communication skills,” which the rumor mill immediately spun as a “no foreigners allowed” policy.
In response, Cui Tianyi set the record straight. “It’s the same deal as an American company in the same space requiring you to work in English—DeepSeek just needs you to be able to work in Chinese,” he clarified. He went on to emphasize, “There is absolutely no rule against hiring foreigners.”
On June 21st, local time, Cui had already posted about the team’s ambitious goals. “As a brand-new department, the DeepSeek Harness group has huge ambitions and a massive workload, so we’re seriously short-staffed. I’m doing interviews every single day and plastering our little ads everywhere… We’ve got three open positions right now: Harness Researcher, Harness Engineer, and Harness Product Manager.” He explained that while the number of vacancies has gone up, the hiring process is identical to the rest of DeepSeek—typically a written test followed by three rounds of interviews, with him acting as the final gatekeeper. All roles are based in Beijing.
Public records show that Cui Tianyi, a post-90s talent, is an alumnus of Zhejiang University—the same alma mater as DeepSeek’s founder, Liang Wenfeng. He previously co-founded TSY Capital, a quantitative trading firm in Hong Kong, where he focused on systematic global equity strategies. This March, he officially joined DeepSeek to lead the brand-new, independent business line called the Agent Harness team.
DeepSeek Harness is an internal coding agent engineering squad dedicated to turning model capabilities into agents that actually get stuff done. Their core mantra is “Model + Harness = Agent,” and right now, the team is going all-in on recruiting.
The job posting for R&D engineers specifies that candidates need at least 2 years of software development experience (though they’re willing to bend this for exceptional people), a bachelor’s degree or higher, and mad skills in using AI Agent tools for development, plus an incredibly fast learning curve. They’re looking for die-hard Agent users who are familiar with the fundamentals of LLMs, including APIs, KV Cache, and the Agent Loop. And, of course, good Chinese communication skills are a must.
Interestingly, the poster really highlights the “good-to-haves.” These include experience with rapid iteration in an AI industry or small team, open-source contributions or personal projects, deep collaboration with researchers or a research background, English communication skills, and any influential competition awards, published papers, or just any out-of-the-ordinary talent that sets you apart.
In the comments section of the recruitment post, one netizen asked, “I heard DeepSeek only hires grads from Tsinghua, Peking, and Zhejiang Universities—is that true?” Cui Tianyi’s reply was a blunt “False.” Another comment pointed out, “The DeepSeek team is still way too mysterious to the outside world, which might be having a tiny impact on recruiting.” Cui agreed, remarking, “You’ve got a point. Staying low-key is a double-edged sword.”
As a top-tier domestic general-purpose model maker, DeepSeek’s recent fundraising moves have been under a glaring spotlight. On June 16th, word on the street was that DeepSeek had completed its first external funding round, pulling in over 50 billion yuan, with a post-investment valuation soaring past 338 billion yuan.
Investment institutions told NUPIAO that in this round, founder Liang Wenfeng likely chipped in around 20 billion yuan, making him the single largest investor. Tencent put in about 10 billion yuan; the CATL system contributed roughly 5 billion yuan, coming from CATL itself and Puquan Capital. NetEase, JD.com, Monolith Management, and IDG Capital each invested around 3 billion yuan, while Loyal Valley Capital and Shixiang Technology each put in about 1.5 billion yuan.
Earlier, on June 3rd, market whispers had disclosed that DeepSeek was aiming to raise about $7 billion in its first round, with a valuation potentially reaching up to $59 billion. Tencent and CATL were pegged to become DeepSeek’s biggest external backers, with NetEase and JD.com also reportedly planning to join the party.
Going back to May 6th, media reports had suggested that the National Integrated Circuit Industry Investment Fund was in talks to lead DeepSeek’s first external round, with a post-investment valuation potentially hitting around $45 billion.
As of press time, DeepSeek has yet to issue any official response to these fundraising rumors.
Even as DeepSeek’s valuation skyrockets, it’s still locked in a fierce domestic AI battle. On one hand, it needs to pump in more cash to expand its computing power and keep its models iterating. On the other, it has to boost pay and benefits to hold onto its top R&D talent.
From the second half of last year to now, at least five key R&D members have confirmed their departures from DeepSeek. The list includes Wang Bing, a core author of the first-gen large language model; Luo Fuli, a key contributor to V3; Guo Daya, a core researcher on R1; plus Wei Haoran, the lead author for the OCR series, and Ruan Chong, a key contributor to multimodal achievements. Most of them have opted to jump ship to internet giants like ByteDance, Tencent, and Xiaomi.
It’s an open secret in the industry that the gap between the supply of AI talent and the hiring demands of expanding companies is massive. A report from Maimai showed that from January to April 2026, AI job postings multiplied by 8.7 times, with their share of all new economy jobs jumping from around 2.78% to 22.03%. “AI” has become the central battleground for spring recruitment. Talent hiring in the new economy is still highly concentrated in first-tier cities. Beijing, Shanghai, Shenzhen, and Hangzhou are the top four hot cities, with Beijing leading the pack with an AI penetration rate of 30.17% for new jobs. That means for every 10 new positions posted, 3 of them are AI-related.
In the ranking of companies with the most new job openings, ByteDance still holds a commanding lead. DJI has leapfrogged Xiaohongshu to grab second place for the first time, with Xiaohongshu now in third. Other major players like Tencent, Meituan, and Ant Group are also firmly at the top of the list. At the same time, companies like MiniMax and Dreame are also breaking into the top 20 hottest employers, proving that smart hardware and AI firms are joining the talent war with fierce momentum.
Against this backdrop, even the famously low-profile DeepSeek has no choice but to hit the gas on hiring.
