AI Hiring Brief · 2026-08-01
32 frontier AI labs · 112 roles added in 24h · agihunt.info/en/jobs
Today at a glance
- OpenAI pivots to harm operations and external affairs: Five of today's 6 additions land in policy-adjacent functions — "Program Manager, Critical Harm Operations," "Cyber Operations Strategist, Critical Harm Operations," "Strategic Narrative and Impact Lead, Intelligence and Investigations," and "US External Affairs Associate, Global Affairs" (Washington, DC) — with "Hardware Platform Security Architect" rounding out a day focused on adversarial-use mitigation rather than product or model work.
- Microsoft AI adds 68 roles in a single day: "Principal Program Manager - Agentic AI," "Principal Group Product Manager - Copilot," and "Software Engineer II and/or Senior Software Engineer - AI Compilers" anchor the technical additions, while nationalization-program internships in Oman, Kuwait, Qatar, and Saudi Arabia indicate structured government-aligned market entry across four Gulf states simultaneously.
- NVIDIA threads silicon, Physical AI, and bio-inference in 11 roles: Three Bengaluru hires in chip design — "SOC Design Engineer," "Power Architecture Engineer," "Senior Packaging Technical Engineer - Hardware" — alongside "Senior Research Scientist - Generative World Models for Autonomous Driving and Physical AI" (Santa Clara) and "Product Manager - BioNeMo Inference" confirm parallel bets across hardware, autonomous systems, and life sciences.
- Cerebras goes all-in on datacenter hardware: All three of today's roles — "Datacenter Electrical Engineer," "Datacenter Architect," and "Systems Debug Engineer- EE" — are physical infrastructure and reliability hires, not research or sales; the company is building hosting capacity around its chips.
- ByteDance Doubao/Seed closes 43 while adding 3: The additions are coordinated world-model product roles — "世界模型产品经理(应用产品方向)" (world model product manager, application products direction) and "世界模型产品经理(模型训练方向)" (world model product manager, model training direction) — plus "大模型自动评测专家(文娱方向)" (large model automated evaluation specialist, entertainment direction); the gap signals active organizational consolidation.
Company moves
OpenAI (753 open; 6 added, 4 closed today; 55 over 7 days): Critical Harm Operations picks up both a program manager and a "Cyber Operations Strategist"; Intelligence & Investigations adds a "Strategic Narrative and Impact Lead"; Global Affairs posts a DC-based external-affairs associate. The only commercial addition is "AWS Specialist Seller, Strategic Pursuits" spanning San Francisco, New York, and Seattle. The vocabulary here — investigations, cyber operations, strategic narrative — borrows from threat-intelligence practice rather than conventional trust-and-safety language.
Google DeepMind (84 open; 5 added, 7 closed today; 18 over 7 days): Two Gemini consumer-app engineers — "Software Engineer, GeminiApp Android and Chrome Integration" (Mountain View) and "Staff Software Engineer, GeminiApp Growth and Discovery" (Bengaluru) — arrive alongside "Head of AI Ecosystem, Kaggle" and "Senior Network Engineer, Robotics," the latter specifically network infrastructure for the robotics program, not research. Net closures outpace additions today.
Microsoft AI (547 open; 68 added, 61 closed today; 207 over 7 days): "Principal Program Manager - Agentic AI" and "Principal Group Product Manager - Copilot" set the product layer; "Software Engineer II and/or Senior Software Engineer - AI Compilers" deepens the platform stack. "Principal Software Engineer, FDE" and "Principal Technical Program Manager - Forward Deployed Engineering" (both UK) extend the forward-deployed engineering model to Europe. Four named government-nationalization internship programs (Oman, Kuwait, Qatar, Saudi Arabia) posted in a single batch indicate structured commercial agreements with host governments rather than routine hiring. Principal engineers land in Bengaluru, Hyderabad, and Noida.
Meta AI (143 open; 2 added, 1 closed today; 14 over 7 days): "Sensor Hardware Engineer," cross-listed across AR/VR, Facebook Reality Labs, and AI Infrastructure (Sunnyvale), pairs with "Research Engineer, Computer Vision" — small but consistent evidence that hardware perception for mixed-reality devices runs alongside core AI research.
NVIDIA (561 open; 11 added, 7 closed today; 41 over 7 days): The Bengaluru cluster — "SOC Design Engineer," "Power Architecture Engineer," "Senior Packaging Technical Engineer - Hardware" — marks India as a genuine chip design site. "Senior Research Scientist - Generative World Models for Autonomous Driving and Physical AI" (Santa Clara) and "Product Manager - BioNeMo Inference" represent the research and productization ends of two separate verticals. "Engineering Manager, AI Compiler Analysis" and "Senior Software Engineer, Networking DGX Cloud" add platform depth.
Apple AIML (339 open; 4 added, 9 closed today; 23 over 7 days): Two "AIML - Data Scientist, Evaluation" roles (Cupertino and Seattle), one "AIML - Sr Machine Learning Engineer, Data and ML Innovation," and one "ML Software Engineer" (both Seattle). Net closures exceed additions; the evaluation concentration continues Apple's pattern of investing in rigorous internal measurement infrastructure.
Cohere (142 open; 4 added, 1 closed today; 10 over 7 days): All four additions are "Business Development Representative" hires — Berlin, Toronto, UK (Spanish-speaking), Dubai — a pure geographic sales expansion with no technical additions today.
Scale AI (211 open; 3 added, 1 closed today; 10 over 7 days): "Senior Mission Software Engineer, Public Sector" spanning Austin, Honolulu, St. Louis, and Washington DC suggests defense or government adjacency; "Staff Applied AI Engineer" lands in London; "Software Engineer, Gen AI" covers San Francisco and New York.
Cerebras (125 open; 3 added, 0 closed today; 11 over 7 days): "Datacenter Electrical Engineer," "Datacenter Architect," and "Systems Debug Engineer- EE" all sit in Hardware Departments or Quality and Reliability — the company is expanding physical compute hosting capacity, not chip research.
ByteDance Doubao/Seed (字节跳动豆包/Seed, 937 open; 3 added, 43 closed today; 9 over 7 days): The three new roles target world-model product ownership, with dedicated product managers for the application layer and the model training layer, alongside an entertainment-direction evaluation specialist. The 43 same-day closures represent the sharpest net contraction in today's dataset.
ModelBest (面壁智能, 135 open; 1 added, 1 closed today; 6 over 7 days) posts a single "政企行业市场营销负责人/市场营销BP" (head of government/enterprise marketing / marketing business partner) in Beijing, formalizing a go-to-market function for the government and enterprise segment. Lovable (63 open; 1 added, 2 closed today) adds "Software Engineer, Platform (Enterprise)" in London. SpaceXAI (195 open; 1 added today) posts "Manager, Facilities Operations" in Data Center, covering Memphis and Southaven, building out mid-south compute operations.
Anthropic (400 open), Perplexity (85), Mistral AI (168), Cursor (119), Together AI (58), and all remaining tracked Chinese labs — Moonshot AI (月之暗面 Kimi), DeepSeek (深度求索), Zhipu AI (智谱 AI), StepFun (阶跃星辰), Meituan LongCat (美团·龙猫), Baidu AI (百度 AI), Tencent Hunyuan (腾讯混元), MiniMax (海螺), 01.AI (零一万物), SenseTime SenseNova (商汤日日新), and Alibaba Tongyi Qwen (阿里通义千问) — posted no daily additions.
What it signals
Microsoft AI's 207 additions over 7 days are the dominant feature of this week. The scale, geographic spread, and functional breadth — compilers, agentic AI, Copilot product management, forward-deployed engineering in Europe, and four simultaneous Gulf government-nationalization programs — indicate a coordinated global expansion rather than reactive headcount filling. The Gulf bloc is particularly precise: Oman Nationalization, Kuwait Nationalization, Erada (Qatar), and Ru'aa (Saudi Arabia) as distinct named programs in a single batch point to contractual arrangements with host governments rather than speculative hires.
NVIDIA's 41 seven-day additions span more strategic axes than any other hardware-adjacent player in the dataset. Silicon design in Bengaluru, Physical AI research in Santa Clara, bio-inference product management, and DGX Cloud networking represent four independent bets being staffed in parallel. For a company trying to position itself as the universal platform layer for AI compute, this breadth is structural rather than opportunistic — each vertical that scales on NVIDIA hardware requires NVIDIA-side product and research depth to match it.
OpenAI's composition today — harm operations, cyber strategy, investigations, external affairs — contrasts with last week's broader pace of 55 additions. The titles borrow from counterintelligence and threat-intelligence vocabulary ("Cyber Operations Strategist," "Strategic Narrative and Impact Lead, Intelligence and Investigations") rather than standard trust-and-safety language, suggesting the organization is standing up an adversarial-use response function with a different operational posture than conventional content moderation. A single day's batch is not conclusive, but the clustering of such titles together points to deliberate organizational design rather than backfill.
ByteDance Doubao/Seed's 43 same-day closures against 3 additions is the China-side story of the day. The 937-role open pool remains the largest in the dataset by a wide margin, but a net contraction of 40 in a single session — combined with the specificity of the three additions (world-model product owners for both the application and training layers, plus an entertainment-direction evaluator) — suggests an internal restructuring aimed at consolidating world-model ownership. The entertainment-direction evaluation specialist alongside dedicated product managers is consistent with ByteDance's core media business as the first intended commercial deployment target for world-model capabilities.