AI Hiring Brief · 2026-10-01

32 frontier AI labs · 53 roles added in 24h · agihunt.info/en/jobs

Today at a glance

Silicon and infrastructure hiring stacks up across the majors: Google DeepMind posted a Silicon Engineer, Power Analysis and a Staff Software Engineer, GenAI Silicon Automation (Paris) on the same day NVIDIA added a PhD Research Intern, AI Accelerator Design and VLSI - 2027 and Cerebras brought on two data-center-focused counsel roles — physical infrastructure keeps pace with model research.
ByteDance runs its sharpest single-day cull of the window: ByteDance Doubao/Seed (字节跳动豆包/Seed) closed 22 roles today against just one addition, nearly triple its net gain of 8 roles over the past 7 days.
OpenAI leans into robotics and international go-to-market: an Engineering Technician, Actuators role under Research sits alongside Growth Lead, Japan and Growth Lead, SEA, plus a Cyber Enablement Lead, Government — hardware, international expansion, and government work all landed in one day.
Apple posts a rare Beijing-based AI role: Machine Learning Engineer - AI, Search & Knowledge Platform is based in Beijing, an uncommon location for an Apple AIML posting.
Chinese labs stay largely quiet again: of the 13 China-based labs tracked, 9 posted zero new roles today and most show flat or near-flat 7-day totals, continuing the gap seen in the previous report.

Company moves

OpenAI (834 open, +11/-13 today, +58 over 7 days) mixed hardware and international signals: Engineering Technician, Actuators under Research points to continued physical-robotics work, while Growth Lead, Japan and Growth Lead, SEA extend its go-to-market footprint in Asia. A Cyber Enablement Lead, Government and Model Policy Manager, National Security round out the government/safety-systems push.

Anthropic (637 open, +6/-8, +54 over 7 days) added Applied AI Architect, International Policy and Applied AI Architects, Partner based in Tokyo, alongside two Compute-team roles — Network Deployment and Maintenance Lead - Data Center Operations and Data Center Global Repairs Program Support — showing enterprise/international expansion running in parallel with data-center operations.

Google DeepMind (92 open, +9/-3, +39 over 7 days) had its busiest day of the window: Senior Research Engineer, Embodied Generalist Agent, Gaming and Research Scientist, Multilingual, Multicultural and Multimodal LLM both in Tokyo, a Research Scientist, Gemini Vision posting across four cities, plus Staff Software Engineer, GenAI Silicon Automation in Paris and Silicon Engineer, Power Analysis in Mountain View — agents, multimodal research, Japan-based research, and silicon all appeared in a single day.

Meta AI (198 open, +4/-3, +27 over 7 days) posted Component Quality Engineer, Optics and Component Quality Engineer, Thermal and Mechanical under AI Infrastructure, plus a Product Quality Engineer, Advanced Manufacturing role spanning Singapore and Taipei — manufacturing quality control for AI hardware — alongside an AI Research Scientist, AI Safety and Security.

NVIDIA (538 open, +6/-3, +17 over 7 days) kept its silicon-recruiting pipeline running with PhD Research Intern, AI Accelerator Design and VLSI - 2027 and Applied Machine Learning Engineer, AI for VLSI Design - New College Grad 2026, both aimed at incoming 2026/2027 graduates.

Apple AIML (296 open, +7/-2, +19 over 7 days) added Machine Learning Engineer - AI, Search & Knowledge Platform in Beijing, a notable location for the team, alongside Character Behavior Developer - Generative AI Experiences SW and Sr Machine Learning Engineer, Proactive — both pointing toward agentic, on-device assistant features.

SpaceXAI (291 open, +4/-5, +17 over 7 days) ran Member of Technical Staff - Evaluation Infrastructure under Engineering alongside Carpenter (Construction) - Memphis and Critical Facilities Technician for its Data Center group in Southaven, MS / Memphis, TN — model evaluation work and physical data-center construction moving together.

Cohere (137 open, +0/-7, +6 over 7 days) closed 7 roles today without a single new posting to offset it.

Scale AI (193 open, +2/-6, +4 over 7 days) added Product Marketing Lead, Public Sector in Washington, DC and Public Sector Lead, Defence & Security in London, continuing to build out its government and defense sales presence.

Cerebras (115 open, +2/0, +7 over 7 days) posted two legal roles specifically for Data Center & Infrastructure — Managing Counsel and Senior Counsel — legal staffing tracking its data-center expansion.

StepFun (阶跃星辰, 231 open, +1/0, +7 over 7 days) added 算力资源规划 (compute resource planning), a role dedicated to planning compute capacity.

ByteDance Doubao/Seed (字节跳动豆包/Seed, 874 open, +1/-22, +8 over 7 days) closed 22 roles today, its heaviest single-day cull in the window, against only one addition: 资深测试开发工程师（AI方向） - Seed 产品研发 (Senior Test Development Engineer, AI direction — Seed product R&D).

The remaining 20 tracked labs — Microsoft AI, Perplexity, Mistral AI, Cursor, Lovable, Together AI, Pika Labs, Aleph Alpha, Genmo, Moonshot AI (月之暗面 Kimi), DeepSeek (深度求索), Zhipu AI (智谱 AI), ModelBest (面壁智能), Meituan LongCat (美团·龙猫), Baidu AI (百度 AI), Tencent Hunyuan (腾讯混元), MiniMax (Hailuo) (MiniMax 海螺), 01.AI (Yi) (零一万物), SenseTime SenseNova (商汤日日新), and Alibaba Tongyi Qwen (阿里通义千问) — posted no new roles today.

What it signals

The global/China hiring gap held again this cycle. Across the 19 global labs tracked, 7-day additions total 268 roles; the 13 China-based labs added just 23 combined over the same window — roughly an 11x difference, in line with the ~10x gap noted in the previous report. Within China, activity is concentrated in a handful of names: ByteDance (8), StepFun (7), MiniMax (3), Baidu (4), and ModelBest (1), while Moonshot AI, DeepSeek, Zhipu AI, Meituan LongCat, Tencent Hunyuan, 01.AI, SenseTime SenseNova, and Alibaba Tongyi Qwen show no net 7-day movement at all.

Physical infrastructure keeps showing up alongside model research rather than behind it. Google DeepMind's silicon roles (Silicon Engineer, Power Analysis; Staff Software Engineer, GenAI Silicon Automation) landed the same day as its embodied-agent and multimodal research postings; NVIDIA's VLSI-focused internships and new-grad roles point the same direction; Cerebras staffed legal counsel specifically for data-center infrastructure; Anthropic and SpaceXAI both posted data-center operations and construction roles. None of this is new by itself, but the breadth — five separate labs posting infra-adjacent roles in a single day — suggests the current hiring wave is as much about building capacity as it is about research headcount.

ByteDance's 22 closures today are the sharpest single-company pruning event of the window, and they echo NVIDIA's 22 closures from the previous day's report almost exactly in scale. Two different companies running a large-scale requisition cleanup on consecutive days, without a matching wave of new postings, reads as routine pipeline hygiene at very large employers rather than a retreat — both companies had healthy 7-day additions (17 for NVIDIA, 8 for ByteDance) even after the cuts. Cohere's day — 7 closures and zero additions — is a smaller-scale version of the same pattern.

OpenAI and Anthropic remain the two largest single-day movers by volume (11 and 6 added respectively, 58 and 54 over 7 days), and their new postings continue to skew toward applied AI and international go-to-market roles — Tokyo and Singapore-based growth and architect positions, plus government-facing work at OpenAI. That combination — enterprise/government expansion layered on top of steady research hiring — has been the consistent shape of both labs' hiring across recent days, with today's postings fitting the same mold rather than marking a shift.