AI Hiring Brief · 2026-07-27

32 frontier AI labs · 4 roles added in 24h · agihunt.info/en/jobs

Today at a glance

Quiet on additions: Across all 32 tracked companies, only four new roles surfaced today — one each from Microsoft AI, SpaceXAI, and two from NVIDIA — marking one of the slower 24-hour windows of the week. The story today is in the seven-day arc and in closures, not fresh openings.
NVIDIA doubles down on silicon architecture: Both of today's NVIDIA additions are chip-and-model-hardware roles — a "Deep Learning Performance Architect" in Shanghai and a "SoC Modelling Architect" across two unspecified locations — signaling continued investment in the full compute stack, from silicon modeling to inference optimization.
SpaceXAI builds out physical data-center infrastructure: A "Critical Facilities Engineer" posted to Memphis, Tennessee under the Data Center department confirms SpaceXAI is staffing the facilities layer of its AI compute build, not just software.
Microsoft AI extends enterprise reach into Latin America: A "Customer Success Account Manager (CSAM)" in Santiago, Chile — while a single post — is notable against the backdrop of 204 Microsoft AI additions over seven days, the highest 7-day count of any tracked company this period.
ByteDance Doubao/Seed leads today's closures: Nineteen roles closed at ByteDance Doubao/Seed with zero additions, continuing a pattern of net-negative days that trims its already-large open pool; the absolute count of 956 open roles still dwarfs every other tracked organization.

Company moves

NVIDIA added 2 and closed 2 today. The new roles land squarely in the hardware-software interface: "Deep Learning Performance Architect" is concerned with squeezing model throughput out of GPU silicon, while "SoC Modelling Architect" operates one level deeper, in the pre-silicon modeling of the chip itself. The Shanghai posting for the performance architect role is the more geographically notable detail — it extends NVIDIA's engineering footprint into mainland China for architecture-level work rather than sales or support. Over seven days NVIDIA has added 47 roles, third-highest among global labs this week.

Microsoft AI added 1 and closed 15 today. The single addition, "Customer Success Account Manager (CSAM)" in Santiago, Chile, extends an enterprise-sales pattern that has dominated Microsoft AI's week: 204 roles added over seven days is almost three times the next-highest global lab. The 15 closures today likely reflect role fills or normal pipeline cleanup rather than a strategic pullback — the 7-day net is still strongly positive. The Santiago hire, specifically in Customer Success Account Management, suggests Microsoft AI is building a Latin American customer success layer to support enterprise deployments of its AI products.

SpaceXAI added 1 with no closures. The "Critical Facilities Engineer" role sits in the Data Center department and is located in Memphis, Tennessee. At 12 roles added over seven days against a 223-role open pool, SpaceXAI's pace is modest but deliberate, and this post reinforces the physical-infrastructure dimension of its AI ambitions — power, cooling, and facilities management for compute clusters, not just the software running on top.

Google DeepMind closed 2 with no additions today; Meta AI also closed 2 with no additions. Lovable and Meituan LongCat (美团·龙猫) each shed 1 role. The remaining companies — including Anthropic, OpenAI, Apple AIML, Perplexity, Cursor, Cohere, Scale AI, Cerebras, and all other tracked Chinese labs — saw no movement in the past 24 hours.

What it signals

The seven-day figures paint a clearer picture than today's quiet surface. Microsoft AI's 204 additions in seven days stands apart from every other tracked company — OpenAI is next at 75, then NVIDIA at 47 and Anthropic at 39. That gap is not noise. Microsoft AI's open pool of 526 and its 7-day velocity suggest an organization in active build-out mode across enterprise sales, customer success, and cloud AI services, while the pure research labs have settled into a steadier cadence. The Santiago CSAM post is one data point, but it fits a pattern of geographic distribution in Microsoft AI's recent additions that the other global labs have not matched.

NVIDIA's hardware-architecture hiring — "SoC Modelling Architect" alongside "Deep Learning Performance Architect" — is a reminder that the compute race is not purely a software story. Modeling how a chip performs before tape-out is expensive, specialized work, and staffing it signals NVIDIA is designing future-generation silicon with AI inference characteristics baked in from the earliest design stages, not retrofitted. The Shanghai location for one of these roles adds a geopolitical dimension worth noting: architecture-level chip engineering in China is precisely the kind of activity that draws regulatory attention, so its visibility in a public job posting is itself informative about how NVIDIA is navigating those constraints.

Among Chinese labs, the 7-day leaders are ModelBest (面壁智能) at 22, Baidu AI at 18, and ByteDance Doubao/Seed at 18 — but ByteDance's 19 closures today push its net negative on the day. ByteDance Doubao/Seed's 956-role open count remains by far the largest of any tracked organization, nearly double OpenAI's 753, but the recurring pattern of large single-day closure batches suggests active role-fill activity or periodic pruning of listings rather than static accumulation. The Chinese labs as a group show lower 7-day velocity than the top global players, with the notable exception of ModelBest, whose 22 additions in seven days against a 133-role pool represents proportionally aggressive hiring.

SpaceXAI's data-center infrastructure post fits a broader theme visible across multiple companies this week: the physical layer — data centers, facilities, power infrastructure — is becoming a staffing priority alongside the model and software layers. As AI compute demands scale, the organizations that own or operate their own compute are moving beyond simply posting ML engineering roles; they are building the operational and engineering teams needed to keep large clusters running at the facilities level. That shift in role composition, from pure software to mixed software-plus-infrastructure, is one of the more durable signals in the aggregate hiring data over the past several weeks.