AI Hiring Brief · 2026-07-25

32 frontier AI labs · 135 roles added in 24h · agihunt.info/en/jobs

Today at a glance

Microsoft AI leads by a wide margin: 46 new roles on the day and 200 over seven days — AI infrastructure for the MAI Superintelligence Team in Zurich and London, agents science in Bangalore and London, Security AI dual openings, and commercial hires spanning Japan, Singapore, Thailand, Australia, and Brazil simultaneously.
Meta makes a chip play: Four of six new Meta AI roles are hardware — "Hardware Engineer, Accelerator Design," "Hardware Engineer, AI Accelerator Module Validation," and two "Silicon Validation Engineer, Infra Silicon" — paired with a post-training research scientist at Meta Superintelligence Labs, linking silicon investment directly to frontier model work.
OpenAI fills the demand side: Six of 13 new roles land in Marketing (paid channels, lifecycle, mid-market, account-based management); "AI Success Engineer - Healthcare & Life Sciences" opens in both San Francisco and New York, treating enterprise vertical deployment as a defined go-to-market motion.
Agent productization is China's shared agenda: Baidu AI opens agent backend, client-side, and security engineering in a single day; ModelBest (面壁智能) adds an agent delivery manager and product lead; Tencent Hunyuan posts an agentic evaluation role for its voice LLM — a coordinated shift from model training to agent deployment visible across multiple labs at once.
Safety acquiring institutional weight: Google DeepMind expands Safety Oversight to two sites; Scale AI clusters five public-sector security roles around Washington DC; Anthropic's "Staff+ Software Engineer, Account Compromise" in London puts account security on a dedicated engineering footing in the UK.

Company moves

Microsoft AI (+46, 7d +200) posted "Member of Technical Staff - Software Engineer (AI infra)- MAI Superintelligence Team" in Zurich and London, a "Principal Applied Scientist- Foundation Models, Agents & Decision Intelligence" in Bangalore, and a "Senior Applied Scientist - Copilot & Agents" in London, distributing agent and infrastructure research to three continents. Commercial roles launched in Japan, Singapore, Thailand, Australia, and Brazil on the same day.

OpenAI (+13, 7d +69) ran a concentrated marketing build: "Senior Marketing Strategist, Paid & New Demand Channels," "Senior Lifecycle Strategist," "Senior Mid-Market & SMB Demand Strategist," and "Senior Account Based Management Strategist" arrived in one sweep alongside "AI Success Engineer - Healthcare & Life Sciences" in two cities. "Software Engineer, Privacy Engineering (Lawful Access)" in Dublin and London marks EU law-enforcement compliance gaining dedicated engineering headcount.

Anthropic (+8, 7d +34) is staffing international deal infrastructure — two "Head of Deal Desk - International" seats in London and Dublin, "Cloud Partner Enablement Lead," "Brand Marketing Lead, Enterprise" — plus "Staff+ Software Engineer, Account Compromise" in London, establishing UK account security as a distinct function.

Google DeepMind (+7, 7d +29) extended robotics evaluation with "Senior Research Engineer, Benchmarking, Robotics," expanded Safety Oversight to Mountain View and London with two matching research engineer postings, and opened a "Researcher, Frontier Strategy and Governance" across four cities (Mountain View, New York, Washington DC, London) — governance now structured as a multi-site function.

Meta AI (+6, 7d +25) posted four hardware roles in one session — accelerator design, module validation, silicon validation ×2 — plus a post-training researcher at Meta Superintelligence Labs, pairing chip-layer investment with frontier model research.

NVIDIA (+8, 7d +43) added chip engineering roles across India and the US: "ASIC Verification Engineer," "Senior Mixed Signal and Analog Circuit Designer," "Senior DFX Engineer - HSIO," "Senior Silicon Reliability Engineer," "Senior ASIC Physical Design Engineer, Netlisting." A "Developer Relations Manager, CAE" adds industrial simulation ecosystem coverage.

Apple AIML (+4, 7d +20) posted "Sr Manager, Siri Agentic Evaluation" in Cambridge and "Software Engineering Manager - Siri Evaluation, Test Development Tooling" in Cupertino alongside two "Applied AI Scientist - Multimodal Intelligence" roles in Seattle and Sunnyvale — Siri agentic evaluation and multimodal research building in parallel at two research centers.

Perplexity (+5, 7d +11) added "AI Strategist, Financial Services" and "AI Strategist, Legal" — vertical enterprise roles positioned between domain consultant and sales engineer — plus a "GTM Strategy & Operations Lead," a visible move from consumer search toward high-value B2B sectors.

Scale AI (+11, 7d +21) clustered five public-sector security roles around Washington DC: "Tech Lead, ARC Team," "Infrastructure Security Engineer, Public Sector," "Security Engineer, Public Sector," "Product Security Engineer, Public Sector," and "Senior Machine Learning Engineer, Public Sector." A "Senior Technician, Robotics" extends Physical AI Operations; a "Sales Enablement Lead" in Dubai opens the Middle East commercially.

Baidu AI (百度 AI, +5, 7d +18) opened "Agent后端研发工程师" (agent backend engineer), "智能体客户端研发工程师" (agent client-side engineer), and "智能体安全工程师" (agent security engineer) in a single day — agent-systems security appearing as a named engineering discipline alongside the core development roles.

Tencent Hunyuan (腾讯混元, +5, 7d +7) added "混元语音大模型 agentic 评测" (agentic evaluation specialist for Hunyuan voice LLM), "混元多模态数据开发工程师" (multimodal data engineer), and "混元多模态大模型算法加速工程师" (multimodal LLM acceleration engineer) — voice agent evaluation and multimodal model efficiency advancing in tandem.

ModelBest (面壁智能, +4, 7d +22) posted "AI 智能体系统交付经理" (agent systems delivery manager), "智能体产品负责人" (agent product lead), and "端侧 AI 业务线市场负责人" (on-device AI business line marketing lead) — one of the few labs treating edge deployment as a commercial product line requiring its own marketing leadership.

Cohere (+4, 7d +11) posted "Engineering Manager, FDE Agentic Platform" for Dubai and Saudi Arabia — senior agentic platform engineering leadership placed in the Middle East — plus "Account Executive Korea" and "Account Executive Japan," filling out the Asia-Pacific commercial layer.

ByteDance Doubao/Seed (字节跳动豆包/Seed) closed 74 roles while adding 2, the largest single-day closure count in today's data. The two additions — a CapCut short-drama agent engineer ("大模型算法工程师（短剧Agent）") and a Doubao product manager for user experience — are narrowly focused consumer product refinements against a backdrop of significant consolidation.

StepFun (阶跃星辰, +2) added "大模型强化学习算法研究员" (RL algorithm researcher for large models), signaling continued core training investment. 01.AI (Yi) (零一万物, +1) posted "大模型与智能体行业销售专家" (LLM and agent industry sales specialist) in Chongqing, extending B2B coverage beyond first-tier cities. Cursor (+2) added a "Product Education Engineer" for customer success and an "Integrated Campaigns, APJ" role targeting Australia. Cerebras (+1) posted a "Director, Data Center Counsel." SpaceXAI (+1) listed a "Trainer - X Money" under Safety in Bastrop, Texas.

Mistral AI, 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), MiniMax (海螺), 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

Microsoft AI's 200 seven-day additions are more than double OpenAI's 69, and unlike a single-front push, the distribution covers AI infrastructure, agents research, security, and commercial expansion at the same time. No other company in today's data is building out all four functions in parallel — the pattern points to a strategy where execution scale, not research specialization, is the competitive lever. The 46 single-day additions come alongside 49 closures, indicating active role turnover rather than pure headcount growth, but the seven-day cumulative makes the direction legible.

The hardware thread is a structural signal. Meta's four chip roles in one session — at a company that has historically been a chip buyer, not a maker — suggest appetite for silicon supply independence. NVIDIA's eight additions, mostly design and verification roles, reflect the upstream pressure driving that appetite: AI compute demand is outpacing reliable supply, and both buyers and makers are staffing for it. Meta's post-training researcher at Superintelligence Labs and its hardware engineers are the same strategic bet viewed from different layers of the stack.

The agent productization wave in China now reads as a phase transition rather than isolated decisions. Baidu, Tencent Hunyuan, ModelBest, and ByteDance all posted agent-scoped roles today — covering backend engineering, client-side development, security, delivery management, product ownership, and evaluation. The same transition is visible in the global data: Microsoft's Copilot & Agents scientists, Cohere's agentic platform engineering manager in the Middle East, Perplexity's vertical AI strategists. The hiring center of gravity is shifting from training foundation models to building the deployment scaffolding around them. ModelBest's on-device AI marketing lead — treating edge inference as a commercial product line rather than a research prototype — adds a dimension not yet visible at peer labs. ByteDance's 74 closures alongside only 2 additions may indicate deliberate scope consolidation into fewer product lines rather than continued broad expansion.

Safety and governance hiring is acquiring structural weight. Google DeepMind's Safety Oversight team expanding to two sites, Anthropic's account security engineering in London, OpenAI's Privacy Engineering (Lawful Access) roles in Dublin and London, and Scale AI's five public-sector security roles clustered around Washington DC are all expressions of the same pressure: as AI systems move into regulated industries and government contracts, safety and compliance infrastructure must scale with the commercial footprint. The roles appearing are no longer advisory or junior — they are staff-level, principal, and team-lead positions with multi-city or multi-site mandates.