A reader wrote to me last month with a familiar puzzle: three job adverts, all titled “DevSecOps Engineer”, describing three different jobs. One wanted a penetration tester who could write Terraform. One wanted a build engineer who could read scanner output. The third, in truth, wanted a compliance analyst with CI/CD vocabulary. All three adverts paid differently, and none of them was wrong about its own team. If you are weighing this career, that confusion is the first thing to understand rather than fight — and it says something honest about where the field stands.
What a DevSecOps role actually is (no two adverts agree)
The field has a confession to make before any career advice: the job title carries no standard scope. There is no official definition of the discipline, no certifying body, and no exam that separates practitioners from spectators — the term circulates in NIST documents and defence guidance, but as a description of posture, not a licence. So every “DevSecOps engineer” advert is really a local arrangement, written by a team solving its own gaps.
Underneath the varying packaging, though, the day-to-day work converges on a recognisable set. You place checks in a delivery pipeline and keep them calibrated — deciding which findings block a merge, which merely record, and which rules have become noise. You handle credentials as a lifecycle problem, from scanning at commit to rotation after exposure. You produce and verify evidence: component inventories, signatures, provenance statements that survive from build to deploy. You run short threat-modeling sessions with developers who would rather ship. And you keep the pipeline itself hardened, because the machinery that builds software is now a target in its own right. That is the craft. Titles are local dialects.
The skills that transfer from where you already are
Nobody is born into this work; everyone arrives sideways, and the three main doors each bring something the others lack.
From software development: you already read code the way scanners wish they could. Your edge is judgement — knowing that a finding is exploitable in this codebase, not in theory — and your gap is usually the pipeline itself. Learn how builds, artefacts, and deploy credentials actually flow, then learn to tune a scanner rather than merely tolerate one. A developer who can cut a rule’s false-positive rate from seventy percent to ten is doing the field’s most valuable, least glamorous work.
From operations and reliability: you already treat systems as things that fail and are observed. The reliability instincts — measure first, automate the response, keep a runbook — map directly onto gate telemetry and compromise rehearsals. Your gap is usually application-level reasoning: what makes code vulnerable, not just what makes systems fall over. A structured primer on the delivery side closes it faster than you expect; this CI/CD pipelines guide covers exactly the machinery you would be defending.
From security: you know the threat models and the catalogue of failure classes. Your gap is flow — the developer’s impatience, the release train’s rhythm, the reason a gate gets overridden at 4:45 on a Friday. Spend time inside a delivery team’s actual cadence before designing controls for it, and the controls you design will survive contact with reality.
Across all three paths, a short list does most of the daily lifting: reading and tuning automated analysis output; threat modeling in the fast format (the forty-five-minute variant, walking STRIDE categories over a single change); enough scripting to glue tools into pipelines; supply-chain mechanics like SBOMs, signing, and provenance verification; and the writing skill to produce a gate policy a human can actually follow. Notice how little of that list is “security knowledge” in the certification-exam sense — most of it is systems fluency plus judgement, which is why the field hires happily from strange backgrounds and why good guidance engineers have ever written tends to age well here.
How teams really staff the work
The arrangement that recurs in healthy organisations is not a department; it is a role inside each delivery team. Security champions — one or two engineers per team — carry protected time, typically a tenth to a fifth of the working week, formally reserved rather than squeezed in after standups. They tune rules, decide exceptions, run the modelling sessions, and stay connected to specialist security staff who handle the deep or novel cases. The pattern works because judgement stays next to the code while expertise stays reachable.
Two career consequences follow. First, the champion seat is the most accessible entry point in the field: you can hold it while remaining a developer or operator, which makes it the natural apprenticeship. Second, dedicated roles do exist — usually titled some variation of DevSecOps or product-security engineer — and they suit people who want to operate across many teams, own the tooling estate, and drive policy. The honest trade-off: dedicated roles risk the exact distance the champion model exists to prevent, so the good ones travel constantly between teams rather than managing from a dashboard.
A week in the champion seat, described plainly: Monday you review the exception list and chase two that have aged; Tuesday you sit with a developer and cut a noisy rule’s false alarms; Wednesday is a forty-five-minute modelling session on a new payments flow; Thursday you verify that last night’s release carried its provenance paperwork; Friday you write — the policy amendment, the metric summary, the runbook paragraph. It is the most cross-functional writing-heavy engineering seat most people ever hold, and people who like variety tend to stay.
Proving DevSecOps skill without a certificate
Since nothing certifies the discipline, hiring managers read evidence instead. The good news is that the evidence portfolio writes itself if you build deliberately:
- Take one open-source project and produce a component inventory for it — then explain in a paragraph what surprised you.
- Wire a scanner into a small pipeline, publish your gate policy for it, and record what you blocked, warned on, and retired after two weeks of data.
- Sign a build and verify the signature at deploy with cosign or an equivalent, and be ready to explain why verification belongs at the deploy step rather than the build step.
- Run the OpenSSF Scorecard on your own repository and fix the three cheapest findings — then repeat once.
- Rehearse a compromise: pick a documented incident where a build action was hijacked and credentials leaked, and write the runbook your team would have needed that morning.
That list demonstrates every skill from the transfer section, costs nothing but evenings, and produces artefacts you can show. Reference reading costs nothing either: the DevSecOps entry of the Ops Atlas gathers the timeline, the control families, and the honest caveats in one place.
The market, honestly
Demand is real but distributed. Because the discipline is a property of teams rather than a headcount, much of the work is embedded inside existing developer and operator jobs — which is precisely why the champion arrangement keeps spreading, and why “DevSecOps engineer” adverts disagree so wildly. The research underneath the field’s culture claims is worth knowing in an interview: the 2022 DORA analysis associated weak security practice with burnout odds 1.4 times higher, and found culture predicted adoption better than tool spending. Teams that understood this tend to be better employers for this kind of work; teams that bought a scanner and called it a programme tend not to be.
The adjacent disciplines borrow the same vocabulary and will appear in your interviews, so a working map helps: how MLOps differs from DevOps shows the pattern extending to machine learning, and the AIOps roadmap for operations teams covers the intelligence layer that increasingly feeds security signals back into operations. My correspondent, for the record, took the second advert — the build engineer’s seat — on the argument that pipeline fluency is the scarcest half of the craft, and six months in she is the person both other adverts were really describing. Start where the pipeline lives; the security depth follows, and it follows faster than any certificate would have carried you.
