A LinkedIn analysis published today by recruiter Sebastian Cuadros overturns the standard narrative about who builds Anthropic Claude: the company behind one of the world's most capable AI models does not look like a research laboratory. It looks like a hyperscaler that materialized in 18 months, assembled almost entirely from senior engineers who spent the previous decade building production systems at Google, Meta, and Amazon — and it is almost completely closed to anyone without a decade of experience.

Cuadros retrieved every LinkedIn profile listing Anthropic as a current employer — 5,306 in total — then isolated the 1,680 in genuine engineering roles and analyzed 7,986 prior job descriptions those engineers had listed. The result is the most detailed public picture yet of who is actually building Claude, and it lands the same week Anthropic is navigating a near-trillion-dollar valuation, a confidential IPO filing, and an ongoing prediction by its own CEO that AI will replace software engineers before the end of this year.

Frontier AI Has Left the Research Phase: What the Workforce Reveals

The deepest signal in the data is not who Anthropic hired. It is what Anthropic needed them for.

Forty percent of the 1,680 engineers have backgrounds in infrastructure — backend systems, distributed computing, databases, or security. Only 3.3 percent have prior experience in reinforcement learning, the branch of machine learning most directly associated with training frontier models. The ratio is roughly 12 to 1.

That gap is not accidental. It reflects a structural shift in where the competitive bottleneck in frontier AI now sits. The most highly compensated role at Anthropic — a TPU Kernel Engineer position advertised at up to $850,000 a year — requires candidates to optimize machine-learning systems running on Google's Tensor Processing Units, working at the level of low-precision inference, high-throughput sampling, and custom collective communication algorithms debugged at assembly code level. That is not a research role. It is a manufacturing efficiency role for AI inference, and it pays $850,000 because per-token inference cost at scale is the competitive lever that determines pricing, margin, and whether Anthropic can keep pace with OpenAI and Google.

In October 2025, Anthropic signed a cloud partnership with Google that gave it access to up to one million TPUs and more than one gigawatt of AI compute capacity coming online in 2026. At that scale, a kernel-level efficiency improvement that shaves a fraction of a millisecond from an inference call translates directly into hundreds of millions of dollars in operating cost. The $850,000 salary is the market's estimate of what that optimization is worth.

Anthropic Built a 1,680-Engineer Organization Almost Overnight

The speed of assembly is striking on its own terms. Only 15 of the engineers now at Anthropic were there before 2021. The team grew to roughly its current size in two concentrated bursts: 686 engineers joined in 2025 alone, roughly tripling the organization's size in a single year, and 455 more had joined by June 2026. The median tenure across the entire engineering team is 10 months. More than half joined within the past year.

In other words, a 1,680-person engineering organization was assembled from scratch in approximately 18 months — and the people assembled are not early-career. The median prior work experience among the team is 12.2 years. The middle 50 percent of engineers arrived with between 8.8 and 16.5 years of professional experience. Engineers with 13 or more years of experience make up 44 percent of the team. Only 50 engineers had fewer than three years of prior experience before joining.

Google Is the Top Feeder, Not OpenAI

The conventional assumption — that Anthropic, founded largely by former OpenAI researchers, recruits primarily from other AI labs — does not match the data.

Google leads all prior employers by a wide margin, with 405 engineers having Google on their CV. Meta follows at 273, then Amazon at 197, Microsoft at 171, Stripe at 124, Apple at 87, Stanford University at 68, DeepMind at 62, Airbnb at 51, and OpenAI at 48. Half of the entire engineering team has worked at a FAANG company at some point. Anthropic also draws disproportionately from firms known for engineering discipline at scale: Stripe, Databricks, Snowflake, Palantir, and Airbnb all appear prominently.

Approximately 94 engineers came directly from a frontier AI lab — OpenAI is the fifth-largest direct source and DeepMind the sixth — but they represent a small fraction of the team. Anthropic is not primarily built from AI lab alumni. It is built from infrastructure engineers who built large production systems at hyperscalers.

The $850,000 Engineer and Why Pay Curves Are So Steep

Beyond the TPU Kernel Engineer role, Anthropic's compensation structure reflects a deliberate bet on experienced infrastructure talent. A cluster infrastructure engineer role in London advertises the equivalent of $440,000 to $655,000 a year and covers the full lifecycle of Anthropic's compute clusters across cloud and owned data centers — configuration, upgrade, decommissioning, fault recovery, and high-bandwidth interconnection. A cybersecurity product engineering manager role pays $405,000 to $485,000.

Levels.fyi data updated as of June 14, 2026, places the median total compensation for software engineers at Anthropic at $710,000, with a reported high of $920,000.

Eighty percent of Anthropic's engineers share a single title: Member of Technical Staff. Former Instagram co-founder and Anthropic CPO Mike Krieger carries the same designation as every former FAANG executive who joined the engineering organization. The title structure is deliberately flattened — seniority, function, and level do not surface through hierarchy labels.

Does Anthropic Hire Junior Engineers?

Almost never through conventional pathways. The 172 engineers with fewer than six years of experience — roughly 10 percent of the total — are not ordinary junior hires. They enter through tightly filtered pipelines: competitive internships at Meta, Google, DeepMind, Jane Street, Two Sigma, and Citadel; AI safety fellowships including MATS, SERI, and Redwood Research; or elite academic credentials. For this cohort, competition rankings and published research carry more weight than years of experience. Nineteen percent of this group hold PhDs — compared to 13.7 percent overall.

Krieger, who co-founded Instagram and joined Anthropic as chief product officer in May 2024, confirmed the pattern on the New York Times Hard Fork podcast: Anthropic has "tended less to hire fresh college grads," citing an absence of a structured internship program and a preference for engineers who can contribute to complex problems immediately.

The schools represented at the top of the organization tell the same story: Stanford leads with 144 alumni in the engineering organization, followed by UC Berkeley at 118, MIT at 80, Carnegie Mellon at 73, Harvard at 42, Cambridge at 39, University of Washington at 36, Waterloo and Cornell at 35 each, Oxford at 33, and Princeton at 32. Stanford, Berkeley, MIT, and Carnegie Mellon together account for roughly a quarter of the entire engineering team.

The Contradiction at the Center of Anthropic's Public Stance

On January 23, 2026, at the World Economic Forum in Davos, Anthropic CEO Dario Amodei told The Economist that AI could be doing "most, maybe all" of what software engineers do within six to 12 months. He also predicted that up to 50 percent of junior white-collar jobs could disappear within one to five years.

Anthropic simultaneously advertised hundreds of engineering roles, hired 686 engineers in 2025, and set its top engineering compensation at $850,000. Amodei introduced caveats — chip manufacturing constraints and training time could slow the transition — and has noted that Anthropic engineers already use Claude to generate code rather than writing it manually, describing that as transformation rather than elimination. Demis Hassabis of Google DeepMind gave AI only 50 percent odds of reaching human-level capability within a decade; Yann LeCun of Meta said large language models will "never achieve human-level intelligence."

The Cuadros data does not resolve the contradiction. It sharpens it. A company whose CEO said AI will replace software engineers hired 686 of them last year, almost none of whom were fresh graduates.

Where the Software Engineering Job Market Actually Stands

The broader labor market data is consistent with the Anthropic picture at the aggregate level but diverges sharply at the entry level. Software developer job postings have risen approximately 15 percent since mid-2025, per Federal Reserve data, and AI and machine learning engineering roles have grown 85 percent year over year. Senior-level postings climbed from 38.8 percent to 43.1 percent of the total IT job mix year over year. Entry-level postings shrank from 8.1 percent to 7.4 percent.

The clearest independent evidence comes from the Stanford Digital Economy Lab. A working paper by economist Erik Brynjolfsson and colleagues, using payroll data from ADP covering millions of workers, found that employment for software developers aged 22 to 25 declined nearly 20 percent from its late 2022 peak — a decline measured after controlling for firm-level factors unrelated to AI. Workers aged 30 and older in the same high-exposure fields saw employment grow six to 12 percent over the same period.

The U.S. Bureau of Labor Statistics projects 15 percent overall growth in software developer employment from 2024 to 2034 — well above the 3 percent average for all occupations. That projection captures total demand, not distribution by career stage. Anthropic itself is a data point in that distribution: it is adding senior engineers as fast as it can build onboarding infrastructure, while not adding junior engineers at all.

Frequently Asked Questions

Does Anthropic hire entry-level or junior software engineers?

Almost never through conventional pathways. The analysis found only 172 engineers with fewer than six years of prior experience across the entire 1,680-person engineering organization. Those who enter at an early career stage typically arrive through elite competitive pipelines: AI safety fellowships, internships at top-tier firms including DeepMind and Jane Street, or exceptional academic credentials from programs like MIT and ETH Zurich. Anthropic CPO Mike Krieger confirmed on the New York Times Hard Fork podcast that the company has "tended less to hire fresh college grads."

Where do Anthropic's engineers come from?

Google is the single largest source, with 405 engineers having Google on their CV. Meta follows at 273, Amazon at 197, Microsoft at 171, and Stripe at 124. Half of the entire engineering organization has worked at a FAANG company at some point. Only 94 engineers came directly from a frontier AI lab — OpenAI and DeepMind rank fifth and sixth respectively among direct prior employers. Anthropic draws far more heavily from infrastructure-focused companies than from AI research organizations.

What is a TPU Kernel Engineer and why does Anthropic pay up to $850K for one?

A Tensor Processing Unit kernel engineer optimizes machine-learning software to run efficiently on Google's custom AI chips, working below standard machine-learning frameworks — tuning low-precision inference, high-throughput sampling, and collective communication algorithms at near assembly-code level. At Anthropic's scale, with access to up to one million Google TPUs, per-kernel efficiency improvements translate directly into lower per-token inference costs, which determine competitive pricing and operating margins. The $850,000 ceiling reflects the scarcity of engineers who can optimize at that level and the direct revenue impact of their work.

Is the software engineering job market shrinking because of AI?

Not overall — but entry-level roles are contracting sharply while senior demand grows. Software developer job postings rose approximately 15 percent since mid-2025, and AI and machine learning engineering roles grew 85 percent year over year. But a Stanford Digital Economy Lab study using ADP payroll data found that employment for software developers aged 22 to 25 declined nearly 20 percent from its late 2022 peak. The U.S. Bureau of Labor Statistics projects 15 percent overall growth in software developer employment through 2034 — but that growth is increasingly concentrated in experienced and specialized roles, not entry-level positions.