AI Hiring Brief · 2026-07-15

32 frontier AI labs · 1902 roles added in 24h · agihunt.info/en/jobs

Today at a glance

Anthropic turns safety into a commercial vertical: Four "Safeguards Enforcement Analyst" roles appeared in a single day — Fraud & Scams, Ban Evasion & Recidivism, Account Takeover & Credential Abuse, and Access Controls & Identity — alongside a "Strategic Account Executive, Cybersecurity" and a "Strategic Partner Development, Product Partnerships - Cybersecurity," positioning enforcement depth as both internal infrastructure and a market differentiator.
Google DeepMind enters the physical stack: A "Research Engineer, Humanoids, DeepMind" and a "Material Science Research Engineer, DeepMind" appeared together in Mountain View — pairing software control for humanoid robots with the study of novel physical materials, well outside the usual ML hiring perimeter.
Microsoft AI deploys globally at volume: 25 additions in one day spanning at least ten countries; the 7-day total of 144 leads all tracked labs, with a single "Director, AI Business & Security Solutions Engineering" posting covering Luxembourg, UK, Switzerland, Ireland, and Belgium simultaneously.
Gulf sovereign AI moves from aspiration to headcount: Scale AI posted "Engineering Manager, UAE" (Dubai), "Engineering Manager, Saudi Arabia" (Riyadh), and "Senior Engineering Manager, Defense" (Doha) on the same day; Cohere's newly visible listings include a "Forward Deployed Engineer, Sovereign AI" and a "Senior Account Executive - KSA" in Riyadh — independently convergent signals from two labs on the same day.
Voice AI emerges as a dedicated engineering track: ByteDance Doubao/Seed's newly tracked pool carries dense clusters of on-device voice model roles including a "C++端到端语音大模型SDK架构师/专家-Data语音" (C++ end-to-end voice LLM SDK architect); Together AI's newly visible listings include a full four-person Voice AI platform team at staff and senior levels.

Company moves

OpenAI (+4 today, +77 over 7 days, −8 closed) posted a "Clean Energy and New Technology Lead" and an "Environmental Lead," both in Scaling — pointing to active energy-permitting work for data center expansion — alongside a "Quantitative Intelligence Analyst" in a newly visible Intelligence & Investigations function.

Anthropic (+9 today, +55 over 7 days) combined its Trust & Safety enforcement sweep with a "Research Engineer, Chip Design RL (Reinforcement Learning)" — applying RL directly to chip design — and "Staff + Sr. Software Engineer, Cloud Inference Launch Engineering," threading three simultaneous bets: safety infrastructure, novel silicon tooling, and inference throughput.

Google DeepMind (+4 today, +39 over 7 days) kept all new roles deeply technical: humanoids and materials science in Mountain View, a "Research Engineer, Pretraining, DeepMind" in New York, and a "Research Scientist, Reinforcement Learning, DeepMind" in London. Its 7-day addition ratio against its current listing count of 87 is the steepest among global labs.

Microsoft AI (+25 today, +144 over 7 days, −41 closed) is rotating its portfolio heavily. New roles span enterprise solution architecture, public sector cloud sales, and AI transformation program management across every major region; the high closure count alongside high additions signals configuration changes rather than retraction.

Meta AI (+5 today, zero closures) added a "Research Scientist, World Modelling" in Menlo Park, a "Software Engineer, Sensors & Computer Vision" for Reality Labs, and a "Business Support Engineer - Meta Business Agents" in Singapore, extending the agentic product's commercial support into Southeast Asia.

NVIDIA (+11 today, +49 over 7 days) posted an "LLM Engineer, Agentic Researcher Platform" alongside silicon design roles — "Senior Architect, SoC Modelling," "DFT Methodology Engineer," and "Senior DFT Power Methodology Engineer" — spanning from the chip design floor to the agentic software stack.

Cohere (130 open; newly tracked) entered with a complete vertical sales structure: dedicated account executives for Energy & Utilities, Manufacturing, Healthcare & Life Sciences, Financial Services, US Public Sector, and KSA, plus "VP Sales, EMEA" in Berlin and a "Head of Sales Enablement." On the technical side, "Software Engineer, Agents & Automations" and "Software Engineer, North for Finance" (naming an agentic finance product called North) point to active agentic platform buildout, while "Forward Deployed Engineer, Sovereign AI" in Ottawa signals delivery capacity for government contracts.

Together AI (61 open; newly tracked) reveals a dedicated Voice AI product team — "Staff Platform Engineer, Voice AI," "Staff Machine Learning Engineer, Voice AI," and their senior counterparts — alongside data center operations, distributed storage, and HPC infrastructure roles, with fall 2026 internships in inference, model shaping, and frontier agents.

Scale AI (182 open; newly tracked) is organized around two expansion axes: US defense and intelligence (including "Strategic Capture Lead, Homeland Security & Federal Law Enforcement" and a public sector engineering cluster in Washington) and Middle East engineering management in Dubai, Riyadh, and Doha, with healthcare agents and robotics as newer commercial verticals.

Cerebras (98 open; newly tracked) assembled four security roles in a single batch — "Principal AI Security Engineer," "Principal Network Security Architect," "Network Security Engineer," and "Hardware / Low Level Security Engineer" — while advancing chip design (ASIC Architect, multiple Design Verification Engineers) and extending data center operations to Finland and Western Canada.

ByteDance Doubao/Seed (字节跳动豆包/Seed; 940 open; newly tracked) is the largest single pool in today's data. The densest cluster is on-device voice AI: roles include "端侧豆包大模型工程专家（嵌入式AI/端到端语音方向）-Data语音" (on-device Doubao LLM engineering expert, embedded AI and voice direction) and evaluation engineers across three cities. A "具身数据引擎产品经理-Seed" (embodied data engine product manager) is the first visible embodied AI signal in this pool.

Tencent Hunyuan (腾讯混元; 61 open; newly tracked) is building an agent runtime — "混元AI Agent Harness Engineer" and "混元Agent强化学习框架工程师" (Hunyuan agent RL framework engineer) — alongside text-to-3D generation ("混元AIGC算法研究员（文生3D）" / Hunyuan AIGC researcher, text-to-3D) and multimodal world model research.

MiniMax (Hailuo) (MiniMax 海螺; 189 open; newly tracked) leads with deep inference work — "AGI推理引擎工程师 — 下一代KVCache存储与缓存加速" (AGI inference engineer, next-gen KVCache storage and cache acceleration) — alongside MaaS platform architecture and an "海外增长运营负责人" (overseas growth operations lead): inference depth, API platform, and international expansion running in parallel.

Alibaba Tongyi Qwen (阿里通义千问; 28 open; newly tracked) posts a "VLA算法工程师" (vision-language-action algorithm engineer) alongside native multimodal world model roles and full pre-training/post-training/agent-system coverage — one of the more structurally complete foundation-model stacks in today's Chinese lab listings.

01.AI (Yi) (零一万物; 49 open; newly tracked) opened two roles in Astana (Kazakhstan) — a content moderation algorithm engineer and a content safety policy specialist — marking a concrete Central Asia deployment footprint, while gaming-specific roles (user profiling, data mining, numerical optimization) and a visual algorithm engineer for the "绿洲" (Oasis) project point to an active secondary product line.

SenseTime SenseNova (商汤日日新; 16 open; +67 added, −51 closed today) shows the sharpest churn in the dataset: a large batch of additions and an even larger batch of closures leave just 16 open roles. Survivors cluster around cloud infrastructure sales ("大装置-云计算销售经理" / 大装置 cloud computing sales manager, multiple cities) and the "元萝卜" (Yuanluobo) overseas e-commerce build-out, with a lone "多模态基模算法研究员" (multimodal foundation model algorithm researcher) as the core-research survivor.

Baidu AI (+5 today, +18 over 7 days, −20 closed) added a "大模型训推框架技术架构师" (LLM training and inference framework technical architect) and a compute sales role in Hangzhou, consistent with a quiet, sustained build in technical depth and commercial capacity.

Aleph Alpha (9 open; newly tracked) posted eight research roles almost entirely in pre-training — "Senior AI Researcher- Pre-training (f/m/d)," "Senior AI R&D Engineer – Pre-training Data (f/m/d)," and a reinforcement learning researcher — consistent with its European sovereign model focus. Pika Labs (10 open) and Genmo (5 open), both newly tracked, lean heavily on research scientists and ML engineers over commercial hires. DeepSeek, Zhipu AI, Mistral AI, and Meituan LongCat posted zero additions today or over the 7-day window; Perplexity, Apple AIML, SpaceXAI, Cursor, Lovable, Moonshot AI, StepFun, and ModelBest each added at most four roles with no directional signal.

What it signals

The clearest structural development today is safety enforcement maturing into a specialized organizational discipline. Anthropic's four simultaneous analyst postings — each scoped to a distinct threat vector — reflect a platform large enough that fraud, ban evasion, credential abuse, and identity manipulation each warrant dedicated operational headcount. The simultaneous launch of cybersecurity sales roles reframes this enforcement depth as a product asset, not just internal compliance. No other lab in today's data has opened that commercial axis; it is worth watching whether others follow as platforms scale.

Physical AI investment is broadening beyond software control. Google DeepMind's pairing of a humanoid controls engineer with a materials scientist on the same day implies work on the physical substrate itself. Alibaba Tongyi Qwen's "VLA算法工程师" and ByteDance's "具身数据引擎产品经理-Seed" suggest Chinese labs are moving in the same direction, though at an earlier organizational stage. NVIDIA's "LLM Engineer, Agentic Researcher Platform" alongside DFT and SoC architecture roles continues to blur the line between hardware supplier and software platform operator.

Geographically, the Gulf cluster in today's data is unusually concentrated. Scale AI's three simultaneous engineering management openings in the UAE, Saudi Arabia, and Qatar — independent of Cohere's KSA sales and sovereign AI engineering hires on the same day — suggest regional sovereign AI commitments have reached the headcount allocation stage, not just the announcement stage. Microsoft AI's single-day spread across Africa, Latin America, and the Asia-Pacific arc reinforces that enterprise AI deployment is now a global staffing operation, not a headquarters-centric one.

Looking at 7-day trend data for established trackers: Microsoft AI at 144 additions leads all labs, followed by OpenAI at 77 and Anthropic at 55. Google DeepMind's 39 additions against 87 total open roles is the highest velocity ratio in the global lab set. Among Chinese labs with meaningful 7-day signals, Baidu AI at 18 additions is the most active, while DeepSeek and Zhipu AI both registered zero over the same window — a divergence that may reflect different product-cycle stages rather than a simple momentum gap, but is notable given both labs' prominence in model releases over the past year.