AI Hiring Brief · 2026-09-24

32 frontier AI labs · 87 roles added in 24h · agihunt.info/en/jobs

Today at a glance

Codex goes multilingual in Europe: OpenAI posted three identical "Applied AI Engineer, Codex" openings today — in Paris, Munich, and Madrid — while its San Francisco team added three Staff+ engineering roles under "Consumer Monetization" (mobile/full-stack, full-stack, and payments & billing), pairing a coding-agent push into continental Europe with ad/monetization infrastructure at home.
NVIDIA hires for a second chip line: alongside a Senior SRAM Engineer, Circuit Design role in Santa Clara, NVIDIA opened a Technical Program Manager - LPU Software and a Senior ASIC Design Verification Engineer - LPU — custom-silicon hiring that sits outside its core GPU roadmap.
Embodied AI keeps growing at ByteDance: Seed added 具身智能数据挖掘工程师（Seed） (embodied-intelligence data-mining engineer) and 机器人电气连接线束工程师 (robot electrical wiring-harness engineer), even as the team closed 11 other roles today — the robotics build-out is expanding while the rest of its req list churns.
Compute infrastructure hiring continues at Anthropic and SpaceXAI: Anthropic added a Capacity Deployment Lead - Data Center Operations and a Data Center Controls Engineer under its Compute org, and nearly all of SpaceXAI's new postings — welder, pipefitter, rigger, material handler — sit at its Memphis/Southaven data-center site.
Microsoft AI goes quiet for a day: zero roles added or closed today, even after a 211-role net gain over the past seven days — still the largest weekly volume of any lab tracked here.

Company moves

OpenAI added 24 and closed 14 (66 over 7 days). Beyond the Codex/Europe and consumer-monetization roles above, it opened trust-and-safety hires (Applied Risk Standards Specialist, Abuse Investigator - Scams & Fraud) and country-lead Strategy & Operations roles for the UK and France — safety, monetization, and international GTM all moving in parallel.

Anthropic added 13 and closed 9 (57 over 7 days). Its compute build-out sits next to an Applied AI Architect, Public Sector and two Applied AI roles for Startups in London, plus an Engineering Manager, AI Observability — infrastructure, public-sector sales, and internal tooling advancing together.

Google DeepMind added 6 and closed 6 (25 over 7 days), evenly turning over its list. New roles include a Staff Research Scientist/Engineer, Agentic Post Training, a Research Scientist, Gemini Safety and Behavior, and a Semiconductor Material Science Research Scientist — agents, safety, and silicon research all represented in a small batch.

Meta AI added 7 and closed 4 (22 over 7 days). Postings split between AI-infrastructure production/hardware roles (Product Quality Engineer, Production Systems Engineer, Network Production Engineer, Hardware) and research (Research Engineer, Safety Evaluation; Research Scientist, Contextual AI and Multimodal Agents) — infra scaling alongside safety and multimodal-agent research.

NVIDIA added 10 and closed 8 (34 over 7 days). Beyond the LPU/SRAM cluster above, it opened a Senior Product Manager — Embedded AI and Next-Gen Computing and AI Computing Development Intern, TensorRT-LLM roles based in Shanghai — custom silicon and China-facing developer roles moving together.

Apple AIML added 4 and closed 2 (12 over 7 days), including an AIML - Distinguished Engineer, Foundation Model and an AIML - Machine Learning Researcher, MLR — modest but senior-weighted research hiring.

SpaceXAI added 6 and closed 5 (33 over 7 days), with the Memphis/Southaven data-center trades cluster described above plus a Corporate Counsel role spanning four US offices.

Baidu AI (百度 AI) added 4 and closed 1 (19 over 7 days). New roles include 产品经理（J106190） (product manager) for its ACG group, 百度百舸-机器学习平台研发工程师（GO方向）（J106189） (machine-learning platform engineer for its Baige AI-cloud platform, Go track), and 搜索推荐策略工程师（J105621） / 内容理解算法工程师（J100150） (search-recommendation strategy engineer / content-understanding algorithm engineer) for MEG.

ByteDance Doubao/Seed (字节跳动豆包/Seed) added 2 and closed 11 (28 over 7 days) — the heaviest single-day closure count in today's data, even as its two new postings extend the robotics cluster above.

ModelBest (面壁智能) added 1 and closed 2 (5 over 7 days); the new role, 评测工程师 - 智能体方向 (evaluation engineer, agent direction) in Beijing, keeps its focus on agent evaluation.

StepFun (阶跃星辰) added 0 today but closed 2 (10 over 7 days) — a quiet day continuing a net drawdown after last week's additions.

Moonshot AI (月之暗面 Kimi) added 1 (国际化合规, international compliance) with no closures (10 over 7 days) — a compliance-specific hire as it expands overseas.

Aleph Alpha is down to a single open role after adding one (Intern/WS to Chief of Staff) and closing one — its posted headcount has effectively collapsed to a skeleton req list.

Smaller movers: Perplexity, Cursor, Cohere, Scale AI, Cerebras, Pika Labs, and 01.AI (Yi) each added one or two roles today — a Berlin Member of Technical Staff (Machine Learning Research Engineer) at Perplexity, a Forward Deployed Strategist at Cursor, a Head of Non-Tech Talent at Cohere, a Software Engineer, Public Sector - New Grad at Scale AI, a Staff Software Engineer, Inference API at Cerebras, a Data Engineer and Growth Engineer at Pika, and a 投融资实习生 (investment/financing intern) at 01.AI — while Together AI closed one role without adding. Microsoft AI, Mistral, Lovable, Genmo, DeepSeek, Zhipu AI, Meituan LongCat, Tencent Hunyuan, MiniMax (Hailuo), SenseTime SenseNova, and Alibaba Tongyi Qwen showed no change today.

What it signals

The seven-day numbers show a stark split in scale: global labs added 513 roles over the past week against 93 for the China-based labs tracked here, and even setting aside Microsoft AI's outsized 211, the remaining global total (302) is still more than three times China's. That gap looks less like a gap in ambition and more like a gap in the breadth of open requisitions — labs like Microsoft AI and OpenAI are running enterprise sales, safety, infrastructure, and consumer-product hiring simultaneously, while several Chinese labs (DeepSeek, Zhipu AI, Alibaba Tongyi Qwen, SenseTime SenseNova) show zero movement over the same week, suggesting headcount plans that are either already filled or paused rather than actively growing.

Where China's labs are moving, it is concentrated and hardware-adjacent. ByteDance's Seed team stands out, both for its 28 additions over seven days and for a robotics/embodied-AI hiring line that has run for multiple days — today's data-mining and wiring-harness roles extend a cluster that already includes reducer/gearbox design and embedded-drive engineering. That reads as a coherent physical-AI bet, distinct from the LLM-only postings still dominant elsewhere in the region.

On the global side, the pattern points toward productization and physical infrastructure more than pure research. OpenAI's Consumer Monetization engineering roles and Codex's simultaneous expansion into Paris, Munich, and Madrid both suggest a company converting research output into revenue-generating product surfaces across multiple markets, not just the US. Anthropic's parallel investment in data-center capacity roles and startup-focused Applied AI hires in London reads similarly — compute scaling and commercial go-to-market advancing together rather than one preceding the other. NVIDIA's LPU-tagged roles, appearing alongside its usual GPU-centric postings, hint at silicon diversification beyond its core product line.

Safety-specific hiring was thin today by comparison — Google DeepMind's Gemini Safety and Behavior researcher and Meta's Research Engineer, Safety Evaluation were the only explicitly safety-labeled additions, a smaller showing than the multi-role safety clusters that have appeared from OpenAI and Anthropic on other days.