Learning roadmaps
The DevOps roadmap, and the two paths beside it
Three sequenced paths into infrastructure work: DevOps Engineer, Cloud Engineer and Site Reliability Engineer. They share a foundation and then diverge. This page is the comparison, so you can pick one before you spend six months on it.
- 3
- Roadmaps
- 25
- Stages
- 3 to 7
- Months
Which one should you follow?
The honest answer is that they overlap more than they differ for the first two months. What separates them is where you go after the foundation, and how quickly you are employable.
| DevOps Engineer | Cloud Engineer | Site Reliability Engineer | |
|---|---|---|---|
| Time | 4 to 6 months | 3 to 5 months | 5 to 7 months |
| Level | Beginner → Advanced | Beginner → Intermediate | Intermediate → Advanced |
| Stages | 10 | 8 | 7 |
| Start here if | You want the broadest path and the most job openings. It covers the whole delivery chain rather than one layer. | You want the fastest route to employable, and you are happy to go deep on one provider before going wide. | You already write code and operate something in production, and you want the reliability and scale side. |
| First role | Junior DevOps Engineer, Platform Engineer, Build Engineer | Cloud Engineer, Cloud Support Engineer, Associate Solutions Architect | Site Reliability Engineer, Production Engineer, Infrastructure Engineer |
| The catch | It is the longest of the three, because breadth costs time. | Narrower than DevOps, so you will meet CI/CD and containers later rather than early. | The worst first path. It assumes the foundations the other two teach. |
What every path starts with
These five come before the roadmaps diverge, and skipping them is the most common reason people stall three months in. None of them is optional, and none needs to be mastered before you move on.
- 1
Linux
Nearly everything you will operate runs on it. Permissions, processes, signals and the filesystem come up on day one and in every interview.
- 2
Git
Every roadmap, every job and every deployment pipeline assumes it. Branching and resolving a conflict without panicking is the bar.
- 3
Networking
DNS, TCP, TLS and HTTP explain most of what breaks. You cannot debug a service you cannot reach without them.
- 4
One language
Python or Go is enough. You need to read code and write automation, not ship a product.
- 5
Containers
Docker is the shared vocabulary of the whole field. Images, layers and volumes come before any orchestration.
The three roadmaps, stage by stage
Every stage in each path, in order. Open a roadmap for the individual topics, the reasoning behind the sequence, and the resources for each step.
DevOps Engineer
10 stages · 4 to 6 months
A complete, sequenced learning path that takes you from Linux and Git fundamentals to containers, Kubernetes, CI/CD, Infrastructure as Code, cloud and observability.
- 1Foundations: OS & Scripting
- 2Version Control & Collaboration
- 3Networking & Security Basics
- 4Containers
- 5Container Orchestration
- 6Infrastructure as Code
- 7CI/CD Pipelines
- 8Cloud Providers
- 9Monitoring & Observability
- 10DevSecOps & Secrets
Cloud Engineer
8 stages · 3 to 5 months
A structured path covering cloud fundamentals, compute, networking, storage, identity, Infrastructure as Code and cloud-native architecture across AWS, Azure and GCP.
- 1Cloud Fundamentals
- 2Choose a Cloud Provider
- 3Compute
- 4Cloud Networking
- 5Storage & Databases
- 6Identity & Security
- 7Infrastructure as Code
- 8Monitoring, Cost & Architecture
Site Reliability Engineer
7 stages · 5 to 7 months
A structured path to Site Reliability Engineering: SLOs and error budgets, observability, incident response, toil reduction, capacity planning, chaos engineering and release engineering.
- 1Engineering Foundations
- 2SLIs, SLOs & Error Budgets
- 3Monitoring & Observability
- 4Incident Response & On-Call
- 5Toil Reduction & Automation
- 6Reliability & Capacity
- 7Release Engineering
How long this actually takes
The durations above assume ten to fifteen hours a week. At five hours a week, double them. The number that matters is not the calendar but the ratio of building to watching: people who deploy something real after each stage finish in the range quoted, and people who only take courses routinely spend a year and still cannot debug a failing pipeline.
You also do not need to finish a roadmap before applying. Most junior postings are satisfied somewhere around the container and CI/CD stages, and the remaining stages are learned faster on the job than off it. Treat the last third as something you grow into rather than a gate.
How to use a roadmap without wasting it
- 1Pick one path and ignore the other two. Switching at month three is the most common way people end up with three half-finished foundations and no depth in any of them.
- 2Build something at the end of every stage, however small. A stage you cannot demonstrate is a stage you have read rather than learned.
- 3Do the stages in order. The sequence is the entire value of a roadmap: Kubernetes before containers, or Terraform before understanding what it is provisioning, is how people end up copying commands they cannot debug.
- 4Keep one repository as you go. By the end it is the portfolio, and it is worth more in an interview than any certificate.
- 5Stop and apply once you reach CI/CD. The last stages are easier to learn while employed, and interviews will teach you faster than another course will.
Common questions
How long does it take to become a DevOps engineer?
Four to six months of consistent study to be interview-ready for a junior role, assuming ten to fifteen hours a week and some prior exposure to programming or IT support. Starting from nothing, closer to nine months is realistic. The variable that matters is not the hours but whether you build things: people who deploy something real progress several times faster than people who only watch courses.
Which roadmap should I start with?
Cloud Engineer if you want to be employable soonest, because it is the narrowest and cloud roles hire at entry level readily. DevOps Engineer if you want the broadest skill set and the most openings, accepting that it takes longer. SRE is the wrong place to begin: it assumes you already write code and have operated something in production, so it is a second or third move rather than a first.
Do I need a degree to get a DevOps job?
No. It is one of the more credential-agnostic parts of software, and hiring is overwhelmingly about what you can demonstrate. What replaces a degree is evidence: a repository with real infrastructure code, a pipeline somebody can look at, and the ability to explain a failure you debugged. A degree helps with visa sponsorship and with some large enterprises, and almost nowhere else.
Do I need to learn AWS, Azure and GCP?
No, and trying to is a common way to waste months. Learn one properly, and the second takes a fraction of the time because the concepts transfer almost entirely. AWS has the largest job market, Azure dominates in enterprises already invested in Microsoft, and GCP is strongest around Kubernetes and data. Pick by the jobs listed where you live rather than by preference.
Are certifications worth it for DevOps?
They help most when you have no professional experience, because they get you past keyword screening. A cloud associate certification and the CKA are the two with real recognition. They do not substitute for building things, and an interviewer will move past the certificate within a minute to ask you what you have actually run.
Can I get a DevOps job without operations experience?
Yes, though rarely as your first job in the industry. The common routes are sideways: from software engineering, from systems administration, or from support. If you are starting cold, target Cloud Support or Junior Cloud Engineer first, which hire without experience and put you next to real infrastructure, then move across within a year or two.
Is the DevOps roadmap different in 2026?
The foundations have not moved: Linux, Git, networking, containers and one cloud remain the base. What has shifted is above them. Platform engineering has absorbed a lot of what was called DevOps, Kubernetes is assumed rather than differentiating, Terraform and OpenTofu are both live options, and being able to work alongside AI tooling is becoming an expectation rather than an advantage.