AI Hiring Brief · 2026-08-04

32 frontier AI labs · 100 roles added in 24h · agihunt.info/en/jobs

AI Hiring Brief · 2026-08-04

32 frontier AI labs · 100 roles added in 24h · agihunt.info/en/jobs

Today at a glance

Microsoft AI's compute-and-clearance push: Microsoft AI added 41 roles in 24 hours (249 over 7 days, the largest figure in the dataset by a wide margin), and a cluster of them are government-facing — "Senior Cloud Solution Architect - CTJ - SECRET," "Senior Cloud Solution Architect - CTJ - TS/SCI," and "Principal Cloud Solution Architect - CTJ - TS/SCI" all sit in Washington, D.C. or Reston, VA, alongside a "Principal AI Security Researcher" in Redmond — a distinct signal of cleared-cloud and security investment layered on top of the routine software-engineering volume.
OpenAI pairs personalization research with datacenter due diligence: Among OpenAI's 9 new roles are "Research Engineer / Research Scientist - Personal AGI, Personalization-Memory" and "Researcher, Multimodal Safety" on the model side, next to "NeoCloud & Colo Due Diligence Lead" and "Environmental Health & Safety Lead - Datacenter" on the infrastructure side — research and physical compute expansion moving in lockstep.
Google DeepMind's bio outlier: Among 7 new DeepMind roles is a "Senior Virologist," an unusual wet-lab science hire for an AI lab, posted alongside more conventional additions like "Forward Deployed Engineer" and a Singapore-based "Technical Lead Manager, Applied AI."
NVIDIA works both ends of the stack: New NVIDIA postings span silicon ("Senior ASIC Design Engineer – Clocks IP") and applied research ("Senior Agentic AI For Data Science Applied Researcher"), consistent with a lab hiring simultaneously for chip design and for agentic application work.
ModelBest doubles down on omni-modal: ModelBest (面壁智能) posted two senior algorithm roles explicitly for omni-modal work — "全模态算法专家/负责人（SFT方向）" (Omni-modal Algorithm Expert/Lead, SFT direction) and "全模态算法专家/负责人（预训练方向）" (Omni-modal Algorithm Expert/Lead, pretraining direction) — plus an infra-focused "多模态训练加速研发工程师-infra" (Multimodal Training Acceleration R&D Engineer, Infra), a coherent push into multimodal foundation-model capability.

Company moves

OpenAI: 9 new roles split between research (Personal AGI/Personalization-Memory, Multimodal Safety) and Scaling-team infrastructure (NeoCloud & Colo Due Diligence Lead, Strategic Operations for Industrial Compute, Environmental Health & Safety Lead), plus a Go To Market pair in Delhi (AI Deployment Engineer, Account Director) — model research, datacenter capacity, and India expansion all advancing at once, against 27 closures today.

Anthropic: 3 new roles — "Safeguards Enforcement Analyst, User Well-being," "Engineering Manager, Search," and "Applied AI Architect, National Security" in Washington, D.C. — a small but pointed mix of trust-and-safety and national-security applied work.

Google DeepMind (谷歌 DeepMind): 7 new roles led by the "Senior Virologist" hire and a Singapore-based applied-AI lead, plus Gemini-side engineering and product roles (Eval Infra, Android/Chrome integrations, Persona and Behavior) — a spread across science, infra, and consumer product surfaces.

Microsoft AI (微软 AI): 41 new roles today, the largest single-day count in the dataset, dominated by cloud/solution architecture and software engineering across the US, India, and a scatter of EMEA and APAC markets, with a concentrated cleared-cloud cluster in the D.C. area and a security-research hire in Redmond.

Meta AI: 3 new roles, including "AI Research Scientist - Meta Superintelligence Lab (Technical Leadership)" and an "AI Research Manager," both in Menlo Park — leadership-level research hiring for its superintelligence effort.

NVIDIA (英伟达): 8 new roles spanning machine learning engineering, developer relations, cloud inference solutions architecture, and ASIC clock-IP design — the breadth from silicon to developer-facing roles is typical of NVIDIA's dual hardware/software hiring motion.

Apple AIML (苹果 AIML): 2 new roles — "AIML Data Operations Team Lead" in Austin and a senior software engineer in Seattle.

SpaceXAI: 2 new roles, a law-enforcement-response analyst in Bastrop, TX and a "Fire Protection Engineer" in the Data Center department in Memphis, TN — continuing the physical-facilities hiring pattern seen in prior editions.

Perplexity: 1 new role, "Legal and Operations Engineering Lead," spanning San Francisco and Palo Alto.

Lovable: 3 new roles, two of them trust-and-safety focused ("Trust & Safety Engineer," "Platform Safety Analyst," both Stockholm) plus a "Staff Software Engineer, Platform (Enterprise)" in London — a safety and enterprise-platform build-out for a coding-assistant product.

Together AI: 1 new role, "IT Engineer" in San Francisco.

Scale AI: 1 new role, "Chief of Staff, Public Sector Engineering" in Washington, D.C., continuing its public-sector engineering focus.

On the closure side without matching additions, Cursor closed 4 roles, Meituan LongCat (美团·龙猫) closed 5, Cohere closed 1, and Cerebras closed 1.

StepFun (阶跃星辰): 4 new roles — "高级网络工程师" (Senior Network Engineer), "Coding评测工程师" (Coding Evaluation Engineer), "Coding系统工程师" (Coding Systems Engineer), and "资深整合营销" (Senior Integrated Marketing) — infrastructure and coding-evaluation hiring alongside a marketing role.

ModelBest (面壁智能): 4 new roles centered on omni-modal algorithms and training infrastructure (detailed above), plus "AI Native 端侧智能解决方案" (AI Native on-device intelligence solutions) in Beijing — on-device deployment paired with core multimodal research.

Baidu AI (百度 AI): 7 new roles, mostly enterprise sales and solutions architecture ("云计算及大模型KA销售" / Cloud & LLM Key Account Sales, "港澳琼桂解决方案架构师" / HK-Macau-Hainan-Guangxi Solutions Architect, "上海解决方案架构师" / Shanghai Solutions Architect, "大客户销售" / Key Account Sales), plus "AI 产品经理（医生助理方向）" (AI Product Manager, Doctor Assistant direction) and "策略算法工程师" (Strategy Algorithm Engineer) — a commercialization-heavy mix with one healthcare-specific product role.

ByteDance Doubao/Seed (字节跳动豆包/Seed): 1 new role, "世界模型数据资源运营专家 - AI数据与安全" (World Model Data Resource Operations Expert, AI Data & Safety), against 28 closures — continued net contraction of the dataset's single largest roster (934 open).

MiniMax (Hailuo, 海螺): 2 new roles, "国内增长投放专家" (Domestic Growth/Ad-buying Expert) and "AI Agent 产品运营实习生" (AI Agent Product Operations Intern) — growth marketing paired with agent-product operations.

01.AI (Yi, 零一万物): 1 new role, "台湾语音实习生" (Taiwan speech/voice intern), a narrow localization-focused hire.

The remaining companies — Mistral AI, Cursor (additions), Cohere (additions), Cerebras (additions), Pika Labs, Aleph Alpha, Genmo, Moonshot AI, DeepSeek, Zhipu AI, Tencent Hunyuan, SenseTime SenseNova, and Alibaba Tongyi Qwen — posted no new roles today.

What it signals

The seven-day numbers make the day's pattern look less like noise and more like a settled trend: Microsoft AI's 249 additions are roughly 5.2 times OpenAI's 48, about 5.5 times NVIDIA's 45, and nearly 11 times Google DeepMind's 23. No other global lab is within range of Microsoft's pace, and the composition of its new roles — cleared cloud architects, a principal AI security researcher, and a long tail of software engineering across Redmond, India, and a dozen other markets — reads as an enterprise- and government-cloud arm scaling largely independently of frontier-model headline hiring.

Global labs collectively added far more over the week than the tracked China-based labs: 452 versus 46, a roughly 10-to-1 gap. That is not simply a story of open-role scale — ByteDance Doubao/Seed alone carries 934 open roles, more than double any single global lab — but of net movement. ByteDance closed 28 more roles today with only 1 addition, extending a multi-day contraction of its own posted roster even as its total base stays the largest in the dataset. Baidu, StepFun, and ModelBest were the only China-based labs adding meaningfully today, and their additions cluster around commercialization (Baidu's sales and solutions-architect roles) and model capability (ModelBest's omni-modal algorithm leads) rather than the infrastructure-operations and cleared-government hiring dominating the Microsoft/OpenAI/SpaceXAI additions.

Within the global cohort, the split between compute build-out and model research keeps showing up side by side rather than as a strict either/or. OpenAI's day paired a personalization-memory research role with a NeoCloud and colocation due-diligence lead and a datacenter environmental-health-and-safety hire; SpaceXAI's only additions were a law-enforcement-response analyst and a fire-protection engineer tied to its data-center footprint. Read together with Microsoft's cleared-cloud cluster, physical infrastructure and government/enterprise access currently account for a disproportionate share of net new roles among the labs with the largest hiring volume, even as smaller counts of frontier-research and safety titles (OpenAI's Multimodal Safety researcher, Anthropic's National Security applied-AI architect, DeepSeek's Senior Virologist at Google DeepMind) continue to appear in low, steady numbers.

Ten of the 32 tracked companies — including Mistral AI, DeepSeek, Zhipu AI, and Alibaba Tongyi Qwen — showed zero movement today, a reminder that hiring activity in this dataset remains concentrated in a handful of labs rather than distributed evenly across the field.