Particle.news

DeepSeek Struggles to Staff 'Harness' Team After Massive Funding

The company’s hiring gap could slow delivery of the middleware that links large language models to agent products.

Overview

  • DeepSeek’s Harness team leader, Cui Tianyi, said on June 22 that the group is severely understaffed and is conducting daily interviews while repeatedly reposting open roles.
  • The company has three long-open role types—Agent Harness researcher, R&D engineer, and product manager—recruited for positions in Hangzhou and Beijing with intern and full-time tracks.
  • DeepSeek says it does not bar hiring foreigners but requires employees to be able to work in Chinese, a language condition it frames like English requirements at U.S. firms.
  • Reportedly following a first external financing round larger than ¥50 billion, the company asked investors to accept a no-poach clause that aims to keep staff from being hired away or spun out.
  • DeepSeek is recruiting simultaneously for MW-to-GW data-center engineering roles in Ulanqab, signaling that it is building compute capacity in parallel with the Harness middleware effort and raising competition for rare cross-disciplinary talent.