DevOps Roadmap for Beginners: What to Learn, in Order

Why this order

Each step uses what the last one gave you. You learn networking on a Linux server, because you need a server to send traffic to. You put your scripts in Git so you have a place to keep them. Then you run an app by hand, and each later tool replaces work you have already done yourself. Docker packages it, CI/CD ships it, and Terraform rebuilds what you first clicked together on AWS.

Roadmaps do not all agree, and you should know where this one takes a side:

• Linux or programming first. roadmap.sh's beginner path and GeeksforGeeks start with a programming language. KodeKloud starts with Linux, as we do. We chose Linux because a beginner gets quick feedback from a shell, and the first programs worth writing here are ones that automate shell work.
• Where Git goes. The milanm roadmap puts Git first, and KodeKloud puts it second. We put it third and keep it small. Nothing much depends on the exact spot, so long as you have it before you write your scripts.
• Where Kubernetes goes. This is the largest disagreement. KodeKloud teaches it before CI/CD and cloud, while roadmap.sh's beginner path leaves it out. We put it last among the technical steps, even though job posts ask for it often. The Kubernetes step gives the full argument, including the case against us.

Our order is a judgment call. No study we found compares learning orders, so these are reasons and not proof.

How long it takes

Our estimate is about 50 weeks at 10 hours a week, across 15 steps. That is arithmetic: we estimated each step and added them up. It is not data, and nobody has timed a group of learners on this path. It does not include your job search.

Published claims range widely, and most come from companies selling a course:

• GeeksforGeeks lays out 13 weeks, which we do not find credible for a true beginner.
• Zero To Mastery advertises 5 months as the average to finish its path. Its own guide says 7-12+ months at up to 10 hours a week. Finishing a course is also not the same as getting hired.
• Spacelift says about six months, assuming you already have basic Linux and networking skills.
• Boot.dev says about 12 months for most beginners.
• Nucamp says 9-18 months at 10-20 hours a week.

These figures assume different hours and different starting points, so do not average them. We found no formal study of how long this takes.

If you can pass a step's test, skip the step. That is how developers and sysadmins go faster. A developer can often test out of Git and much of the scripting. A sysadmin can often test out of Linux, networking and running an app by hand. Try the test first, and be honest about the result.

What we left out and why

More tools make a roadmap look complete, but each one takes time from the skills you need first. These are all real and useful. They are not needed for a first pass:

• Ansible automates setup across many servers that already exist. It matters when you manage a fleet of long-lived servers, and many employers use it. Your scripts, containers and Terraform cover the same ground on this path. You can pick it up on the job.
• Jenkins is an automation server that teams often use for CI/CD. It matters if your employer already runs it, and many do. The concepts carry over from GitHub Actions, which needs no server of your own.
• Service mesh, such as Istio, adds a proxy that handles traffic between services. It matters when a company has many services and needs consistent security and routing between them. One small app has no such problem.
• Serverless, such as AWS Lambda, runs your code without you managing servers. It matters for small jobs that run on an event or that see uneven traffic. We left it out because you need to understand servers first.
• Multi-cloud means running on more than one provider. It matters at some large companies and to some regulated ones. Learn one cloud well, since most of the ideas carry over when a job asks for another.

How to use AI while you learn

This is a rule for every step. It is not a step of its own.

• Use it as a tutor. Ask it to explain an error, quiz you on a topic, or point out what your plan is missing. Try the problem yourself first, or you will skip the part that teaches you.
• Verify before you run. AI tools can be confidently wrong, and a wrong command can delete real data. Check anything that deletes or changes things against the official docs. Run it on a machine you can rebuild.
• Be able to explain any command. If you cannot say what each part does, do not run it. An interviewer, or a future outage, will ask you the same question.

Working developers seem to take a similar view. In the Stack Overflow 2026 survey, 48.0% of those who answered trust AI when they can easily verify the output. Only 6.6% trust it with important work decisions. That survey covers developers in general, not beginners or DevOps engineers alone.

We found no credible data on how AI has changed junior DevOps hiring. Plenty of people make confident claims in both directions, but they are opinions.

Who this is for

We assume you can use a computer and have never run a server. You do not need a degree, a programming background or a particular job today.

If you are further along, use the step tests to find your starting point. If you only want one tool, go straight to its page. The later pages assume the earlier ones, so you may need to go back for a step or two.