AI Hiring Brief · 2026-07-17

32 frontier AI labs · 95 roles added in 24h · agihunt.info/en/jobs

Today at a glance

Microsoft AI leads the day on hardware depth: At 27 new additions — the highest single-day count across the field — Microsoft AI's new postings include "Sr. Hardware Engineer TPM," "Principal Design Verification Engineer" (Silicon Engineering, India), "Principal Supercomputing Operations Software Engineer," and "Principal AI Accelerator Tools Development Engineer," confirming a sustained push to build proprietary silicon and supercomputing infrastructure alongside its software stack.
SpaceXAI opens a Central Asian language data sweep: Five remote "AI Tutor" roles in Sinhala, Tajik, Mongolian, Kazakh, and Kyrgyz appeared in a single batch alongside two data center power-engineering roles in Memphis — a company running physical infrastructure expansion and low-resource language data acquisition on parallel tracks.
NVIDIA consolidates its autonomous-vehicle perception layer: "Senior Radar Perception Engineer, Obstacle Foundation Models - Autonomous Vehicles" (Santa Clara), "Senior Machine Learning Engineer - Automotive" (Austria), and "Senior Software Engineer, Multi-Sensor Fusion" (Shanghai) landed simultaneously across three countries — a coordinated sub-team build, not scattered headcount.
Security governance is becoming a named function: Cursor posted "Regional Security Manager," Scale AI posted "Head of International Security" in London, and Google DeepMind posted "UK and Europe National Security Lead" — three organizations treating security as a dedicated operational hire on the same date.
Cohere's additions read as pre-IPO compliance build-out: All three of today's new roles target the controllership and audit layer — "Internal Audit and SOX Compliance Manager," "Finance Systems and Automation Manager," and "Assistant Controller" — against a backdrop of 141 seven-day additions.

Company moves

Microsoft AI posted 27 additions today (154 over seven days, closing 46). Hardware and silicon roles are the defining feature: "Sr. Hardware Engineer TPM," "Principal AI Accelerator Tools Development Engineer" spanning California, Oregon, and Washington, and "Principal Design Verification Engineer" in Silicon Engineering in India all appeared alongside routine software engineering in Redmond. International distribution extended to Germany, Singapore, the Netherlands, Ireland, Mexico, and Israel — Microsoft AI is absorbing headcount across geographies at a pace no other organization in this dataset matches this week.

NVIDIA added 12 today (53 over seven days). Three autonomous-vehicle roles — radar perception, multi-sensor fusion, and automotive machine learning — form a coherent perception-stack sub-team spread across Santa Clara, Shanghai, and Austria. The concurrent "Relational Foundation Model Engineer, Modern Data Stack" is an outlier worth watching: NVIDIA applying foundation-model thinking to structured enterprise data is a category extension that has not appeared prominently before.

SpaceXAI added 8 today (15 over seven days, no closures). "Electrical Reliability Engineer (Data Center Infrastructure)" and "Power Systems Engineer (Data Center Infrastructure)" in Memphis anchor a self-build data center push; the five Central Asian and South Asian AI Tutor roles are all remote. The language selection — Tajik, Kazakh, Kyrgyz, Mongolian, Sinhala — clusters around markets where English-language training data is thin, suggesting a deliberate low-resource capability investment.

Anthropic added 4 today (52 over seven days, closing 8). "Product Manager, Safeguards (Verticals)" is structurally notable: Safeguards is being treated as a vertical-facing product line, not just an internal safety layer. "Field Marketing Manager, Public Sector" pairs with it. Two finance roles — "Finance Systems Engineer, Revenue" and "Data Science, Finance & Strategy" — signal that revenue has scaled to a point requiring dedicated systems and analytics.

Google DeepMind (谷歌 DeepMind) added 4 today (28 over seven days). "Research Scientist, Multimodal Alignment, Safety and Fairness, DeepMind" spans Mountain View, New York, and Kirkland. "UK and Europe National Security Lead, DeepMind" in London is rare for a research lab — DeepMind is formalizing a government-relations and security-clearance function, likely driven by UK and EU government contracting requirements. "Senior Staff Data Scientist, GeminiApp, DeepMind" points to continued investment in Gemini's consumer-product data layer.

Scale AI added 7 today (194 over seven days, closing 2). "AI Advisory Principal" and "AI Advisory Consultant" in enterprise sales position Scale as an AI transformation consultancy alongside its data-labeling business. "Head of International Security" in London and "DevOps Engineer, Infrastructure & Security" in Washington, DC place security infrastructure at the UK and US government nexus simultaneously.

Cohere added 3 today (141 over seven days, no closures). The three roles — "Internal Audit and SOX Compliance Manager," "Finance Systems and Automation Manager," and "Assistant Controller" — are precisely the controllership layer that companies assemble before a public offering. SOX compliance readiness and revenue-systems automation at this pace signal an IPO preparation timeline in motion.

Cursor added 3 today (10 over seven days). "Deal Desk Manager, APJ" in Singapore and "Regional Director, Geo Enterprise" in San Francisco form the standard enterprise sales maturation pair — deal-desk process in Asia-Pacific, senior field-sales leadership in the US. "Regional Security Manager" in New York adds the operational security credential that large enterprise customers increasingly require.

OpenAI added 9 today (59 over seven days, closing 8), holding 716 open roles. "Data Center Compute Infrastructure" and "BIM Coordinator / BIM Designer" in Scaling reflect continued physical build-out. "AI Success Engineer - EDU" in Singapore is one of the few customer success roles to appear in Asia-Pacific this week. "Academic Research Partnerships & Programs Lead" suggests OpenAI is formalizing university relationships at the program level.

MiniMax (MiniMax 海螺) added 2 today (191 over seven days). "IDC选型与运营工程师" (data center site selection and operations engineer, Beijing/Shanghai) is a role that only appears when a company is actively evaluating and operating owned infrastructure rather than relying on cloud. The 7-day figure of 191 is the third-largest of any organization in this dataset.

StepFun (阶跃星辰) added 3 today (9 over seven days, closing 7). "AI数据处理分析工程师" (AI data processing and analysis engineer), "大模型销售运营（北京上海）" (large model sales operations), and "AI产品运营经理" (AI product operations manager) combine data pipeline, sales process, and product operations — the triad of a lab shifting from R&D toward commercial deployment.

Smaller additions: Together AI posted "Staff Software Engineer, Inference / Compute Infrastructure Engineering"; Cerebras added two infrastructure and testing roles in Sunnyvale and India; Moonshot AI (月之暗面 Kimi) added a partnership manager and a "高性能分布式存储工程师" (high-performance distributed storage engineer); ModelBest (面壁智能) added one "端侧智能软件开发工程师" (on-device intelligent software engineer); and Meituan LongCat (美团·龙猫) posted one "AI native招聘专员" (AI-native recruiter). Companies with no additions today include DeepSeek (深度求索), Zhipu AI (智谱 AI), Baidu AI (百度 AI), ByteDance Doubao/Seed (字节跳动豆包/Seed), Tencent Hunyuan (腾讯混元), Alibaba Tongyi Qwen (阿里通义千问), Mistral AI, Meta AI, Lovable, and Pika Labs.

What it signals

The institutionalization of security stands out as the day's clearest cross-company pattern. Three organizations posted operationally-framed security leadership roles — not red-team researchers or security engineers in the traditional sense, but people responsible for government contracts, regional compliance frameworks, and international security programs. Anthropic's "Product Manager, Safeguards (Verticals)" reframes this further: safety is being packaged as a vertical-specific product, not merely an engineering discipline or a policy response. The convergence suggests that regulatory pressure in the US, UK, and EU has crossed a threshold where it generates real organizational charts, not just white papers.

Microsoft AI's 154 seven-day additions place it at roughly three times the pace of the next-busiest global lab. The character of those additions — silicon design verification, supercomputing operations, AI accelerator tooling — represents hiring that takes 12 to 24 months to fully onboard and cannot be wound down and restarted cheaply. The sustained pace implies a funded, multi-year hardware roadmap rather than speculative exploration. NVIDIA's simultaneous build-out in automotive perception — spanning three countries and targeting the obstacle-modeling and sensor-fusion layers specifically — follows the same logic: these are not generalist hires but specialists who slot into a predetermined architecture.

On the Chinese side, the dominant story this week is not today's quiet but the accumulated seven-day figures: ByteDance Doubao/Seed at 977 additions, MiniMax at 191, SenseTime SenseNova at 67, Tencent Hunyuan at 61. None of these posted new roles today, which likely reflects batch-upload practices more than actual hiring pauses. What the weekly totals do show is that the largest Chinese labs are operating at a scale comparable to or exceeding the global leaders when measured over the full window — ByteDance Doubao/Seed's 977 seven-day figure exceeds every global organization tracked here by a substantial margin. The day-level signal from MiniMax's data center site-selection role and StepFun's sales and operations triad suggests the commercialization layer is being built in parallel with the technical one, following a pattern familiar from earlier waves of Chinese internet-to-AI company transitions.