What 134 DevOps and SRE Job Postings Actually Ask For

134 open roles whose title names DevOps, SRE, platform engineering or infrastructure engineering, taken from 41 public company job boards and checked against 42 terms. Managers and program managers are not in it.

Each figure is the share of postings using a term at least once, so repetition inside one posting counts for nothing. At this sample size the ± column matters more than usual: intervals run to roughly eight points, which means neighbouring rows are not meaningfully different. Read the bands, not the ranks.

A named product reaches further here than in any earlier study. AWS is number 2 of 42 at 75.4%. Every other study we have run is below, and two of them already had a named product near the top — so this is a difference of degree, not a reversal:

Infrastructure hiring names its stack more heavily than product or sales hiring does. It does not name it more heavily than security hiring, which the 212-posting security study already showed.

A practice still beats every one of them. automation leads at 81.3%, above 75.4% for AWS and 70.1% for Kubernetes, and scalability (72.4%) outranks Kubernetes too. The pattern from the earlier studies is dented here, not overturned.

The generic word beats the branded one, again. Observability is named in 63.4% of postings against Grafana 27.6%, Prometheus 22.4% and Datadog 17.9%. That is the same shape as SIEM against Splunk in the security study, where the generic term led by 4 times. Carry both forms on a resume.

The previous generation of tooling has almost gone. Jenkins appears in 3% of these postings, Chef 3.7%, and Puppet in 1 of the 134. Against that, Terraform is at 62.7%, ArgoCD 24.6% and GitHub Actions 19.4%. GitHub Actions is named roughly 6 times as often as Jenkins.

SRE vocabulary is real vocabulary. Toil appears in 26.9% and SLO, SLA or SLI in 35.1% — both ahead of several household tool names. Postmortems reach 12.7% and chaos engineering 6.7%.

Want this for one posting rather than the aggregate? Put your resume and that posting through the free checker and it names the terms the posting uses that you do not.

Where the postings came from. The Greenhouse job board API, published by companies so their own careers pages can render their listings. Nothing scraped, nothing private. The collector and the analyzer are both in the repository, so these numbers can be reproduced against the same endpoint.

Which postings counted. Title match on DevOps, SRE, site reliability, platform engineer or infrastructure engineer. Managers, directors, technical program managers and SkillBridge interns are excluded — twelve postings — because a posting for someone who runs the team asks for different things than a posting for someone on it.

Duplicates. One requisition advertised in several cities returns several records. Company, normalized title and content length together decide whether two are the same posting; the rule is shared with the other studies so the sample sizes stay comparable.

One term needed care. A bare match on “Go” also catches “go to market” and “go above and beyond”, which overstated it by about four and a half points. It is matched case-sensitively with a negative lookahead, the same handling the software engineering study uses. Every other term was checked loose against strict and came back identical — including Helm, which carries no “at the helm” contamination in this corpus.

What this cannot tell you. 41 companies, one ATS vendor, mostly US and mostly venture-funded. Not a random sample of infrastructure jobs. Terms outside the 42-term list were not counted, so nothing here is evidence that a missing term does not matter — no certification was measured at all. And it is one day's snapshot of open roles, 2026-09-09, not a trend line.