AI Hiring Brief · 2026-09-22

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

Today at a glance

Microsoft's enterprise-scale req volume keeps outpacing everyone else: Microsoft AI posted just 1 new role today (a Software Engineer II – Full Stack in Redmond) against 15 closures, but its 7-day total of 239 additions is by far the largest in the dataset — more than four times OpenAI's 52 and roughly double the combined 7-day total of every Chinese lab tracked here.
SpaceXAI doubles down on data-center power infrastructure: 9 of SpaceXAI's 10 new postings today sit under the Data Center department in Southaven/Memphis — Electrical Engineer (Power Generation), Turbine Technician, Plant Technician, Fluids Engineer, Controls Engineer, and a Manager, Engineering (Power Generation & Utilities) — a clear signal of physical buildout around power generation rather than software hiring, alongside one consumer-facing Creator & Community Program Lead — Grok Imagine.
OpenAI splits between consumer hardware and European sales: three new Consumer Devices roles — Operating Systems Engineer, On-Device Inference; Operating Systems Engineer, Linux Kernel; and Operating Systems Engineer, Connectivity & Networking — point at continued device/OS engineering, while Strategic Partnerships Manager postings for Germany and France mark fresh international go-to-market investment in Europe.
NVIDIA leans further into autonomous vehicles: new postings include Senior Machine Learning Engineer, Perception - Autonomous Driving; Senior Software Control Integration Engineer - Autonomous Vehicles; and Senior Software Engineer, AV Planner - Autonomous Vehicles (Shanghai) — a concentrated push into robotics/AV alongside an NIM Solution Architect role for enterprise AI deployment.
StepFun (阶跃星辰) keeps building agent infrastructure: five new roles today include Agent Harness 产品经理 (Agent Harness Product Manager), 全栈研发工程师（Harness） (Full-Stack Engineer, Harness), and Agent Runtime开发工程师 (Agent Runtime Development Engineer) — a continuation of the agent-tooling investment visible in prior days, plus a 【基础模型】战略研究专家 (Foundation Model Strategic Research Expert) role.

Company moves

OpenAI added 10 roles and closed 16 today (52 over 7 days). Beyond the Consumer Devices OS trio and the Germany/France partnerships hires noted above, it also posted a Machine Learning Engineer, Monetization AI/ML and a Manager, Applied AI Architects in London — a mix of device engineering, monetization, and applied-AI GTM staffing.

Anthropic added 7 and closed 2 (46 over 7 days). New roles skew toward finance and sales infrastructure — Corporate Finance & Strategy (Strategic Transactions), Senior Manager IT SOX, a Public Sector Sales Lead in Paris, and a Head of APAC GTM Enablement in Singapore — alongside Applied AI Architect and Customer Success roles in Mumbai and Bangalore, suggesting a deliberate India go-to-market push.

Google DeepMind added 3 and closed 10 (24 over 7 days). New postings span responsible AI research in London, a Staff Software Engineer for Gemini App growth in Bengaluru, and a Tech Lead Manager for Frontier and AGI Security — research, growth, and safety staffed in the same batch.

Microsoft AI added just 1 role today but closed 15, against the week's largest 7-day total of 239 — see above.

Meta AI added nothing today but closed 4 roles, against 13 additions over the past 7 days — a quiet day dominated by requisition pruning rather than new listings.

NVIDIA (英伟达) added 6 and closed 5 (29 over 7 days). Alongside the autonomous-vehicle cluster above, it posted a Developer Relations Manager – CAE and CFD and a Developer Relations Manager, Higher Education and Research, continuing to build out its developer-ecosystem and enterprise-simulation footprint.

Apple AIML added 1 role, a Software Engineer Manager: Agentic Evaluation Platform in Cupertino (22 over 7 days, 6 closed) — evaluation infrastructure for agentic systems.

SpaceXAI added 10 and closed 1 (27 over 7 days) — the Memphis power-generation buildout described above is the dominant theme.

Perplexity added 1 role, an Enterprise Customer Support Specialist (remote US), against 8 additions over 7 days and no closures today.

Cerebras added 3 roles today (10 over 7 days, no closures): an AI Datacenter Infra Engineer plus a Distributed Systems Security Engineer and a Software Supply Chain Security Engineer, both in Sunnyvale — infrastructure security hiring alongside its usual compute build-out.

Baidu AI (百度 AI) added 2 roles: an 智能体算法工程师 (Agent Algorithm Engineer) and an 能源电力解决方案架构师 (Energy & Power Solutions Architect), both in Beijing — agent modeling paired with an energy-sector solutions push.

ByteDance Doubao/Seed (字节跳动豆包/Seed) added 3 but closed 20 today, the heaviest closure count of the day, against 32 additions over 7 days — the highest 7-day figure among Chinese labs. New roles include a VLM大模型数据运营专家（基础模型） (VLM Foundation Model Data Operations Expert), an Edge通用智能模型研究员 (Edge General Intelligence Model Researcher), and an AI异构计算优化专家 (AI Heterogeneous Compute Optimization Expert) — edge deployment and compute efficiency work continuing even as the broader req list is pruned hard.

ModelBest (面壁智能) closed 4 roles today with no new postings, against 8 additions over the past 7 days — a pruning day with no visible new hiring push.

01.AI (Yi, 零一万物) added 1 role, an AI 全栈产品工程实习生 (AI Full-Stack Product Engineering Intern) in Beijing, against 8 additions over 7 days.

Mistral AI, Cursor, Lovable, Cohere, Together AI, Scale AI, Pika Labs, Aleph Alpha, Genmo, Moonshot AI (月之暗面 Kimi), DeepSeek (深度求索), Zhipu AI (智谱 AI), Meituan LongCat (美团·龙猫), Tencent Hunyuan (腾讯混元), MiniMax (Hailuo, MiniMax 海螺), SenseTime SenseNova (商汤日日新), and Alibaba Tongyi Qwen (阿里通义千问) showed no net change today.

What it signals

The single largest number in this dataset is Microsoft AI's 239 additions over 7 days — a figure so far ahead of the rest of the field (OpenAI's 52, Anthropic's 46, NVIDIA's 29) that it reads less like organic growth and more like a distinct hiring mode: enterprise and solution-engineering requisitions opened in bulk, then closed off just as quickly (15 closed today alone). That volatility, high adds paired with high closes, suggests Microsoft is actively reshaping a large sales and solutions organization rather than steadily scaling headcount.

Physical infrastructure is showing up across multiple labs at once. SpaceXAI's Memphis cluster is entirely power-generation and facilities roles — turbine technicians, controls engineers, fluids engineers — which points to data-center energy buildout as a bottleneck worth staffing directly rather than outsourcing. NVIDIA's parallel push into autonomous-vehicle perception and planning roles, including one specifically based in Shanghai, and Cerebras's addition of distributed-systems and supply-chain security engineers alongside its usual infra roles, both suggest that as frontier compute and robotics ambitions scale, the hiring is following the hardware and the supply chain, not just the models.

Comparing the two regions on 7-day volume: global labs added roughly 500 roles combined over the past week, while the Chinese labs tracked here added roughly 105 — even after including Microsoft's outsized total, the ex-Microsoft global figure still outpaces China's total by more than 2-to-1. Within China, the activity is concentrated in just two labs: ByteDance Doubao/Seed (32 over 7 days) and StepFun (26), both oriented toward agent and edge-model infrastructure, while DeepSeek, Zhipu AI, SenseTime SenseNova, and Alibaba Tongyi Qwen show zero net movement over the same window — either a genuine hiring pause or simply no refreshed postings during this period.

Agent tooling is the one theme that spans both regions today without translation needed: StepFun's Agent Harness and Agent Runtime roles mirror the kind of agent-infrastructure investment that's been visible in prior days, and Baidu's own Agent Algorithm Engineer posting suggests the "agent harness" layer, the plumbing that lets a model act rather than just answer, is becoming a hiring priority independent of which lab or country is doing the staffing.