How to become a DevOps engineer: the order that works
Most people who want this job get lost in the tool list — a hundred logos, no order. The order is the whole thing. Learn the layers in the sequence they build on each other and every new tool has somewhere to stand. Here is that sequence, how much code it really needs, which certificates are worth the money, and what the market pays.
New to cloud? CAMPUX is a free, build-first course. Start here →
To become a DevOps engineer, build skills in this order: Linux and the command line, Git, one scripting language, containers, CI/CD pipelines, infrastructure as code, and one cloud platform. Prove it with a portfolio project that ships a containerized app through an automated pipeline. Everything below is the long version of that sentence.
What you are actually aiming at
A DevOps engineer keeps the road between a developer's laptop and production short, fast and hard to break. You automate the build, the tests and the deploy so a change goes out without anyone hand-editing a server at midnight. You describe infrastructure in files instead of clicking through a console, so it can be rebuilt identically. And you watch what you shipped, so you notice trouble before a customer does.
Notice what that job assumes. It sits on top of software other people wrote, servers that run Linux, and a cloud account someone is paying for. You cannot automate a deploy you do not understand, so you learn the ground before the automation. That ordering is not a preference — it is why people stall. They start at Kubernetes because it is the word in the job ad, bounce off it, and conclude they are not cut out for this. They were fine. They just started three floors up.
The skills, in the order to learn them
- Linux and the command line. Servers run Linux, containers run Linux, your pipeline runs on Linux. Get comfortable moving around a filesystem, reading logs, managing permissions and processes, and editing files without a mouse. This is the floor everything else stands on.
- Git. Every automated pipeline triggers off a commit, so version control is not optional plumbing — it is the thing that starts the machine. Learn branching, merging, pull requests, and how to read a history.
- One scripting language. Bash for gluing commands together, Python for anything with logic in it. You are not writing an application; you are writing the small programs that automate the boring parts. Pick one, get fluent, then read the other.
- Containers. Docker first — what an image is, how a container differs from a virtual machine, how to build and run one. Kubernetes after, and only after, because it orchestrates a thing you need to understand first.
- CI/CD. A pipeline that builds, tests and deploys on a commit. This is the centre of the job, and the first thing an interviewer will ask you to describe.
- Infrastructure as code. Terraform or Bicep. Declaring what should exist so the environment is reproducible and reviewable rather than assembled by hand.
- One cloud platform. Deep in one beats shallow in three. The shape transfers when you need the second.
The tool list is intimidating because it is written as a list of products, and products are interchangeable. There is a category called "CI/CD" and it does not matter enormously whether you learned it in GitHub Actions or Azure Pipelines or GitLab, because the concepts — triggers, stages, artifacts, environments, secrets, approvals — move with you. Learn one tool per category well enough to explain the category, and treat the rest as dialects. A candidate who can explain why a pipeline stage exists is more employable than one who has skimmed six tools and can configure none of them under questioning.
One pipeline you built and can defend beats a checklist of twenty tools you have watched videos about.
How much code, really
Enough to automate, not enough to ship a product. Coding for DevOps is writing the small programs that remove manual work: a script that tags a hundred resources, a step that checks something before a deploy, a job that parses an API response and fails the build when a value is wrong. It is not building an application with a user interface, and the interview will not ask you to invert a binary tree.
Bash or Python first?
Take Bash first if you are starting from nothing, because it is the language of the shell you are already living in and the payoff is immediate. Take Python next, and expect it to become the one you reach for most. The working rule once you have both is simple.
Reach for Bash when the task is a short sequence of shell commands — run this, check that, glue a couple of command-line tools together in a pipeline step. Reach for Python the moment the task grows real logic: loops with conditions, an API call, JSON parsing, proper error handling, or anything you will reuse, maintain or test. The clearest tell is length. If a Bash script is sprawling past twenty or thirty lines, or filling with nested logic, that is the script asking to be rewritten in Python. Learning where that line sits is itself a DevOps skill.
And you learn this by automating something, not by working through tutorials. Pick a task you actually do by hand and remove it. The tutorial teaches syntax; the task teaches judgment, and only one of those shows up in an interview.
Do certifications help?
They help at the start and stop mattering later. A certificate gets a résumé past a keyword filter and tells a recruiter you have done the reading. It does not prove you can build a pipeline, and nobody senior is paid more for holding more of them. Treat one as a door-opener and put the rest of your hours into the project.
| Certification | Focus | Difficulty | Résumé value |
|---|---|---|---|
| Microsoft AZ-400 — DevOps Engineer Expert | CI/CD, release management and DevOps practice on Azure | Hard, and it has a prerequisite certification | High for Azure and Microsoft-shop roles |
| CKA — Certified Kubernetes Administrator | Running and operating Kubernetes clusters | Hard; a fully hands-on timed lab | High and vendor-neutral wherever Kubernetes runs |
| HashiCorp Terraform Associate | Infrastructure-as-code fundamentals | Moderate; the most approachable here | Good, cheap, broadly recognised |
| AWS DevOps Engineer Professional | CI/CD, automation and monitoring on AWS | Hard; professional level, broad scope | High for AWS-centric roles |
AZ-400 is an expert certification and you cannot take it cold. Microsoft requires you to hold either the Azure Administrator Associate or the Azure Developer Associate first. Since AZ-204 retired on 31 July 2026 and took the Developer Associate with it, AZ-104 is the only route still open to a new candidate — people who already hold the developer credential still satisfy the requirement. Plan the administrator exam first and budget for two exams, not one.
If you are already an Azure administrator
This is the shortest path into the role and the one I see work most often, because you already have the ground floor. The mindset shift comes before the skills: an administrator is measured on tickets closed and systems kept up, and a DevOps engineer is measured on how rarely anyone has to touch anything by hand. You stop being the person who does the change and become the person who builds the thing that does the change.
The skill delta, in the order that works: Git fluency beyond clone and commit — branches, pull requests, reviewing a diff. One scripting language written properly, with functions and error handling rather than a wall of commands. Infrastructure as code, Bicep first if you are Azure-native and Terraform after, since the second is easy once you have the first. And CI/CD, with federated credentials rather than a stored secret from day one, because doing it the insecure way first is a habit you then have to unlearn.
A realistic timeline, and avoiding tutorial hell
Six months is realistic at ten to fifteen hours a week from a standing start with some IT background; nine to twelve if you are starting further back or have less time. Roughly: a month on Linux and Git, a month on scripting, six weeks on containers, six weeks on pipelines and infrastructure as code, then the rest on the cloud platform and the project.
The failure mode is tutorial hell — an endless queue of courses that each feel productive and none of which leave you able to do anything unaided. The exit is to stop consuming at the point where you can follow along, and start building something that was not in the video. The moment you have to solve a problem the tutorial did not cover is the moment you begin learning, and it is uncomfortable enough that most people retreat to another course instead. Do not.
The project that gets callbacks
One project, done properly, outperforms a list of tools every time. Build a small containerized application and ship it through an automated pipeline into a cloud environment you defined in code. That single artifact demonstrates the whole stack in the order you learned it: Git triggers the pipeline, the pipeline builds a container, infrastructure as code creates what it deploys onto, and monitoring proves it is alive.
Put it in a public repository with a README that says what you built, which decisions you made, and what you would change. Then be able to walk someone through it out loud. The interviewer's question is always some version of "tell me about something you built", and a candidate who can answer it concretely is in a different bracket from one who lists technologies.
The market and what it pays
DevOps sits among the better-paid infrastructure roles, because the skills combine software and operations and the supply of people who genuinely have both is thin. In the US market, mid-level roles commonly land somewhere around $110k–$150k base with senior and platform-leaning roles well above that, and total compensation at equity-paying employers runs considerably higher. Read those as directional; the number moves with region, employer tier and level far more than with the job title.
Two patterns worth knowing. Remote roles trend slightly below comparable on-site ones and are far more competitive, so early on a net that includes hybrid is pragmatic. And the premium skills are the ones that carry accountability rather than the ones with the newest logos: Kubernetes at real scale, security in the pipeline, and cost control all pay better than knowing one more CI tool. If the underlying question is whether the field is worth entering at all, the demand picture is in are Azure cloud engineers in demand.
Questions people also ask
How do I become a DevOps engineer?
Learn the layers in dependency order: Linux and the command line, Git, one scripting language, containers, CI/CD, infrastructure as code, and one cloud platform. Then build one project that ships a containerized app through an automated pipeline into infrastructure you defined in code, put it in a public repository, and be able to walk someone through the decisions. Six months is realistic at ten to fifteen hours a week.
What should I learn first in DevOps?
Linux and the command line, then Git. Both are the floor everything else stands on: containers run Linux, pipelines run on Linux, and every pipeline triggers off a commit. Starting at Kubernetes because it appears in job ads is the most common reason people stall and conclude the field is not for them.
Do you need to code for DevOps?
You need to write code, but not to be a software engineer. It is automation glue: a script that tags resources, a pipeline step that checks something before a deploy, a job that parses an API response. Fluency in one shell language plus enough Python to handle logic and error handling is the working bar, and you build it by automating real tasks rather than by working through tutorials.
Should I learn Bash or Python first for DevOps?
Bash first if you are starting from nothing, because it is the language of the shell you already work in and the payoff is immediate. Python next, and it will become the one you reach for most. The rule afterwards: Bash for a short sequence of shell commands, Python the moment there is real logic, an API call, JSON parsing or anything you will maintain. A Bash script sprawling past thirty lines is asking to be rewritten in Python.
Is a DevOps certification worth it?
At the start, yes, as a door-opener that clears keyword filters and shows you did the reading. Later, no — nobody senior is paid more for holding more of them. Pick one that matches your stack, then spend the remaining hours on a project you can defend, because that is what decides the interview.
What are the prerequisites for AZ-400?
Microsoft requires you to already hold either the Azure Administrator Associate or the Azure Developer Associate before you can earn the DevOps Engineer Expert certification. Because AZ-204 retired on 31 July 2026 and took the Developer Associate with it, AZ-104 is the only route still open to new candidates, though existing developer-credential holders still qualify. Budget for two exams.
How long does it take to become a DevOps engineer?
Around six months at ten to fifteen hours a week if you already have some IT background, and nine to twelve months from further back or with less time available. Roughly a month each on Linux and Git and on scripting, six weeks on containers, six weeks on pipelines and infrastructure as code, and the remainder on a cloud platform and the portfolio project.
Can an Azure administrator move into DevOps?
It is the shortest path in, because the ground floor is already there. The shift is as much mindset as skill: an administrator is measured on tickets closed, a DevOps engineer on how rarely anyone touches anything by hand. The skill delta is Git fluency beyond clone and commit, a scripting language written properly with functions and error handling, infrastructure as code, and CI/CD with federated credentials rather than stored secrets.