AI Hiring Brief · 2026-10-04

32 frontier AI labs · 33 roles added in 24h · agihunt.info/en/jobs

AI Hiring Brief · 2026-10-04

32 frontier AI labs · 33 roles added in 24h

Today at a glance

Chinese labs prune, not hire: ByteDance Doubao/Seed (字节跳动豆包/Seed) closed 10 postings today with zero additions, Baidu AI (百度 AI) closed 4, and Meituan LongCat (美团·龙猫) closed 1 — the day's entire closure volume among China-based labs came with no replacement hiring.
NVIDIA doubles down on robotics and agentic multimodal: two Senior Deep Learning Scientist, Multimodal Agentic RL postings landed alongside a Solutions Architect, Industrial Robotics and a Senior DFT Methodology Engineer, spanning research, go-to-market, and chip validation in one day.
Anthropic builds out trust & safety infrastructure: Program Manager, Safeguards Policy, Enforcement, and Threat Intelligence appeared alongside two separate Security Audit & Controls, Security GRC listings.
SpaceXAI's Memphis data-center buildout continues: Optical Engineer, Commissioning Engineer (Electrical), and Rack Design Engineer all tie to the same Memphis/Southaven data-center corridor.
OpenAI pairs hardware marketing with APAC growth: an Integrated Marketing Manager, Hardware in San Francisco landed the same day as a Total Rewards Business Partner, APAC and an APAC Vendor Lead, Ads, both based in Singapore.

Company moves

OpenAI (828 open, +7/-8 today, +58/7d) added hardware marketing (Integrated Marketing Manager, Hardware) and APAC go-to-market roles (APAC Vendor Lead, Ads; Total Rewards Business Partner, APAC) in Singapore, plus two Applied AI engineering seats (Software Engineer, Enterprise Controls; Software Engineer, OpenAI Presence) — hardware and international expansion moving in parallel.

Anthropic (640 open, +5/-3, +53/7d) clustered around safety and security: Program Manager, Safeguards Policy, Enforcement, and Threat Intelligence plus two Security Audit & Controls, Security GRC roles, alongside a Head of Recruiting, Communications, Brand, & Marketing and a brand Program Manager.

Google DeepMind (谷歌 DeepMind) (98 open, +2/-0, +33/7d) added two Technical Program Manager roles — one for "Scientific Intelligence for Gemini" and one for Compute Management — pointing to growing program-management scaffolding around Gemini research and compute allocation.

Meta AI (205 open, +3/-4, +36/7d) posted a Security Engineering Manager and a Security Engineer tied explicitly to Applied AI/AI Infrastructure, plus an ASIC Engineer, Foundry and IP role split across Sunnyvale and Austin.

NVIDIA (英伟达) (553 open, +4/-0, +29/7d) added two Senior Deep Learning Scientist, Multimodal Agentic RL openings, a Solutions Architect, Industrial Robotics, and a Senior DFT Methodology Engineer — agentic multimodal research, robotics go-to-market, and silicon validation advancing together.

Apple AIML (299 open, +3/-2, +19/7d) added a Machine Learning Scientist for AppleCare demand planning, an AIML - Quality Engineering Manager, Responsible AI and Safety, and a Machine Learning Scientist for GenAI & ML Frameworks in New York.

SpaceXAI (301 open, +3/-1, +21/7d) posted three Data Center roles — Optical Engineer, Commissioning Engineer (Electrical), Rack Design Engineer — concentrated around its Memphis/Southaven campus.

Perplexity (127 open, +3/-0, +7/7d) added a Director of Events and Experiences, a Global Workplace Business Operations Lead, and a Product Manager, Platform — operations and platform-product hires rather than research roles.

Cursor (132 open, +1/-0, +12/7d) added a single Field Marketing Manager, Healthcare / SLED (remote), a vertical-market marketing push.

Cerebras (117 open, +1/-0, +8/7d) added a Senior Research Engineer, Inference ML, continuing its inference-focused research hiring.

Pika Labs (15 open, +1/-0, +1/7d) added a Product Designer.

Baidu AI (百度 AI) (471 open, 0 added, -4 today, +2/7d) closed four postings with no replacements, a pruning day despite a mildly positive weekly net.

ByteDance Doubao/Seed (字节跳动豆包/Seed) (883 open, 0 added, -10 today, +4/7d) closed ten roles today — the largest single-day closure count — even as its 7-day net stays marginally positive.

Meituan LongCat (美团·龙猫) (68 open, 0 added, -1 today, 0/7d) closed one role, flat otherwise for the week.

The remaining companies were unchanged today: Microsoft AI, Mistral AI, Lovable, Cohere, Together AI, Scale AI, Aleph Alpha, Genmo, Moonshot AI (月之暗面 Kimi), DeepSeek (深度求索), Zhipu AI (智谱 AI), StepFun (阶跃星辰), ModelBest (面壁智能), Tencent Hunyuan (腾讯混元), MiniMax (Hailuo) (MiniMax 海螺), 01.AI (Yi) (零一万物), SenseTime SenseNova (商汤日日新), and Alibaba Tongyi Qwen (阿里通义千问) posted no new roles and closed none.

What it signals

The gap between global and China-based labs' weekly hiring pace is stark and consistent. OpenAI's 7-day net (+58) leads the dataset, followed by Anthropic (+53), Meta AI (+36), Google DeepMind (+33), NVIDIA (+29), SpaceXAI (+21), Apple AIML (+19), and Cursor (+12) — all comfortably in double digits. Among the 13 China-based labs tracked, the strongest 7-day net is just +4, shared by ByteDance Doubao/Seed and StepFun, with Baidu AI and MiniMax (Hailuo) at +2, ModelBest at +1, and eight labs — Moonshot AI, DeepSeek, Zhipu AI, Meituan LongCat, Tencent Hunyuan, 01.AI (Yi), SenseTime SenseNova, and Alibaba Tongyi Qwen — flat at zero net change for the week. This isn't a story of paused Chinese AI hiring so much as a different headcount posture: ByteDance Doubao/Seed and Baidu AI carry large standing postings pools (883 and 471 open respectively) but thin net movement, and today's closures at both — plus Meituan LongCat — suggest roles are being refreshed or reallocated rather than grown.

Infrastructure and custom silicon continue to draw hiring across multiple labs at once rather than from a single company's buildout cycle. SpaceXAI's three new Memphis-area data-center engineering roles, Meta AI's ASIC Engineer, Foundry and IP posting, and NVIDIA's Senior DFT Methodology Engineer alongside its industrial-robotics and multimodal-agentic-RL hires all landed in the same window — physical infrastructure and chip-adjacent work remain a steady demand signal independent of any one lab's product cycle.

Safety and security staffing is tracking alongside, not behind, capability hiring. Anthropic's safeguards and security-GRC cluster, Meta's security-engineering roles explicitly scoped to Applied AI and AI Infrastructure, and Apple's Quality Engineering Manager for Responsible AI and Safety all appeared in the same 24-hour window — three labs treating trust-and-safety staffing as a parallel track rather than an afterthought.

Smaller, product-stage companies show a different posture. Perplexity, Cursor, Cerebras, and Pika Labs each added modestly this week, but their new roles — Director of Events and Experiences, Field Marketing Manager, Product Designer — lean toward commercialization and go-to-market rather than research headcount, consistent with labs further along in productizing existing models rather than scaling frontier research teams.