TLDR
Fifty steps across ten stages, in the order they actually happen. Helpdesk 2015, sysadmin 2017, first cloud job February 2020. Six of the ten stages have labs attached, so you can stop reading and go run the thing instead. Work through it, do not read it once.
When to Use
You are in helpdesk, support, sysadmin or development and you want a cloud role. If you already work in cloud and want to move up, this is the wrong page. Start with Cloud Skills 2027.
🗺️ The ten stages at a glance
The order matters more than any individual step. Find where you are, start there.
| Stage | What it decides | Steps | Labs | Skip it if |
|---|---|---|---|---|
| 1. Where to start | Which cloud, and whether you build a habit | 1-5 | 4 | You already deploy weekly in one cloud |
| 2. What to learn | Whether the rest ever makes sense | 6-10 | 8 | You can explain a route table and a role assignment |
| 3. Where to practise | Whether you can fix things, not just build them | 11-15 | 4 | You have broken and repaired your own environment |
| 4. Certifications | How much time you lose to exams | 16-20 | none | Your three job ads do not name a certificate |
| 5. Projects that count | Whether you get interviews at all | 21-25 | 6 | You have three finished, deployed, talkable projects |
| 6. Your resume | Whether anyone reads past line three | 26-30 | none | Never skip |
| 7. Your LinkedIn | Whether anyone finds you | 31-35 | none | Never skip |
| 8. Finding the jobs | Interview volume | 36-40 | none | You already have interviews booked |
| 9. The interview | Whether you convert them | 41-45 | 5 | Never skip |
| 10. The offer, and after | Whether the first job compounds | 46-50 | 4 | You have not had an offer yet |
Gotcha
The single most common failure is doing stage 4 first and stage 5 never. Twelve months in, that leaves you with a certificate, no interviews, and no idea why. Certifications clear a filter. Projects get you the interview.
🧭 Stage 1. Where to start
The decisions that cost six months if you get them wrong.
| # | Step | Why it matters |
|---|---|---|
| 1 | Pick one cloud, not three | Concepts transfer, so the choice matters far less than sticking to it. People who learn all three learn none |
| 2 | Check three local job ads before you pick | Whichever cloud those three name is your cloud. Ten minutes, and it overrules every roadmap online including mine |
| 3 | Open a free account today, not after the course | You cannot learn this by reading. Every cloud gives enough free credit to build something real |
| 4 | Set a spending alert on day one | Before you deploy anything. Every provider lets you, almost nobody does, and it is behind every scary bill story |
| 5 | Block two hours, three times a week, same time | Six hours a week for a year beats a heroic month followed by nothing |
Where the budget setting actually lives, for step 4:
| Cloud | Path | Set it to |
|---|---|---|
| Azure | Cost Management + Billing > Budgets > Add | A monthly cap you would not mind losing, alerts at 50, 80 and 100 percent |
| AWS | Billing and Cost Management > Budgets > Create budget | The same, plus a zero spend budget to catch anything leaving the free tier |
| Billing > Budgets and alerts > Create budget | The same, scoped to your one learning project |
Labs for this stage. Deploy something on day one, before you understand it.
Cloudlearn.ioCreate your first virtual machine in AzureYour first deployed resource. Do this before you understand it.labs.cloudlearn.io
Cloudlearn.ioDeploy your first website with Azure CLIA live site from the command line, not the portal.labs.cloudlearn.io
Cloudlearn.ioDeploy an EC2 instance in AWSThe AWS equivalent, if your three job ads said AWS.labs.cloudlearn.io
Cloudlearn.ioAzure CLI essentialsThe commands you will use every day for the next five years.labs.cloudlearn.io
Pro Tip
If you already work in IT, your current job is the best lab you will ever have access to. Real systems, real users, real consequences. Use it.
📚 Stage 2. What to learn
In this order. Skipping ahead is why cloud feels impossible.
| # | Step | Why it matters |
|---|---|---|
| 6 | Networking first, before anything else | How traffic gets somewhere, why it does not arrive, how to tell the difference. Everything sits on top of this |
| 7 | Identity: who can do what, and where | Users, groups, roles, permissions, scope. The second pillar, and where most real incidents come from |
| 8 | One scripting language, badly is fine | PowerShell, Bash or Python. You are not becoming a developer. Read a script, change it, run it without fear |
| 9 | Linux basics, even on a Windows path | Filesystem, logs, editing a file, permissions. Non-negotiable in 2027 whichever cloud you chose |
| 10 | Read infrastructure as code before you write it | Open a Terraform or Bicep file someone else wrote and work out what it does. Writing from blank comes much later |
Labs for this stage. Networking, then identity, then infrastructure as code.
Cloudlearn.ioCreate and configure VNets and subnetsAddress space and subnets. Start here, everything sits on it.labs.cloudlearn.io
Cloudlearn.ioImplement NSGs and ASGsAllow and deny rules. Where most access problems actually live.labs.cloudlearn.io
Cloudlearn.ioConnect virtual networks with peeringTwo networks talking privately, and what peering does not do.labs.cloudlearn.io
Cloudlearn.ioIntroduction to Amazon VPCThe same ideas in AWS vocabulary.labs.cloudlearn.io
Cloudlearn.ioEntra ID: create users, groups, assign rolesWho can do what, and why you assign to groups not people.labs.cloudlearn.io
Cloudlearn.ioConditional access and MFA in Entra IDThe controls every organisation asks about in interviews.labs.cloudlearn.io
Cloudlearn.ioWrite and deploy your first Bicep templateRead and deploy a template. Reading is the day-one skill.labs.cloudlearn.io
Cloudlearn.ioIntroduction to Terraform on AzureThe other side of infrastructure as code.labs.cloudlearn.io
🧪 Stage 3. Where to practise
| # | Step | Why it matters |
|---|---|---|
| 11 | Use the free tier until it genuinely stops being enough | More capable than people assume. You will not outgrow it in six months |
| 12 | Build in your own account, not a simulator | A sandbox that cannot break teaches you the happy path. The happy path is not the job |
| 13 | Break something on purpose every week | Delete a rule, revoke a permission, change a setting. Watch what fails, then fix it. Highest value habit here |
| 14 | Delete everything you build, then rebuild it | The second build is where the learning is. The first one you followed along, the second you decided |
| 15 | Practise fixing what you did not create | Debugging someone else's work is closer to the real job than anything you build from scratch |
Labs for this stage. The destroy and rebuild cycle, plus the guardrails.
Cloudlearn.ioTerraform workflow: validate, plan, apply, destroyThe full cycle including destroy, which is the half people skip.labs.cloudlearn.io
Cloudlearn.ioTerraform state fundamentals and inspectionWhat state is and why it breaks. This one bites everybody.labs.cloudlearn.io
Cloudlearn.ioBicep conditions, loops and what-if deploymentsWhat-if shows you the damage before you do it. Catch it before you deploy it.labs.cloudlearn.io
Cloudlearn.ioAzure resource locksThe guardrail that stops you deleting something expensive.labs.cloudlearn.io
Gotcha
Deallocate or delete resources when you finish a session. A stopped virtual machine still bills for its disks and its public IP. This catches nearly everyone once.
🎓 Stage 4. Certifications
The industry sells this stage to you backwards.
| # | Step | Why it matters |
|---|---|---|
| 16 | Take one foundational certificate, then stop | AZ-900 or the AWS equivalent. It gives you vocabulary and a structure to hang things on. That is its job |
| 17 | Understand what a certificate actually does | It clears an automated filter. It does not get you into the room and has never closed an interview |
| 18 | Never take two in a row without building in between | Back-to-back exams are the most comfortable procrastination available, because they feel like progress and have a score |
| 19 | Put the certificate near the bottom of your resume | Under the things you have operated. Top of the page is for what you have run |
| 20 | Skip it entirely if your three job ads do not name it | Some markets care a lot, some not at all. Your three ads from step 2 already told you which |
No labs on this stage, deliberately. If you are studying for an exam, the labs in stages 2 and 5 teach you more than another practice test.
🛠️ Stage 5. Projects that count
The stage that gets you hired, and the one most people skip.
| # | Step | Why it matters |
|---|---|---|
| 21 | Three finished projects beats nine started | Nothing counts until it runs. Four things at seventy percent is zero things |
| 22 | Finish one before you start the next | The urge to start something new at the difficult part is what you are training against |
| 23 | Write down what broke and how you fixed it | As it happens, not afterwards. This document becomes your interview answers |
| 24 | Deploy it, do not just build it | Reachable, running, with the mistakes that only appear when something is live |
| 25 | Talk about each one for ten minutes without notes | The real test. If you cannot, it was a tutorial, not a project |
Labs for this stage. End to end rather than single service, which is what makes them worth discussing in an interview.
Build a serverless contact form on AzureSmall, finished and end to end. A real project rather than a tutorial.cloudlearn.io
Cloudlearn.ioContainerise and deploy a multi-container app to Container AppsMore than one moving part, which is what makes it worth discussing.labs.cloudlearn.io
Cloudlearn.ioDeploy a full environment with BicepNetworking, a VM and dependencies in one template. A system, not a resource.labs.cloudlearn.io
Cloudlearn.ioWrite your first GitHub Actions workflowAutomated deployment. The line on your resume that says you ship.labs.cloudlearn.io
Cloudlearn.ioDeploy containerised apps to Azure with GitHub ActionsBuild, push and deploy without touching a portal.labs.cloudlearn.io
Cloudlearn.ioCreate and deploy a REST API with API GatewayAn API you can call from anywhere, on the AWS side.labs.cloudlearn.io
📄 Stage 6. Your resume
| # | Step | Why it matters |
|---|---|---|
| 26 | Lead with what you have run, not a summary paragraph | Nobody reads the summary. The first three lines are the most valuable space on the page |
| 27 | Outcomes, not a technology list | "Azure, Terraform, Kubernetes" tells a hiring manager nothing. What did you do, and what happened |
| 28 | Every line should be safe to be questioned on | A resume is a list of things you are inviting someone to ask about for ten minutes each |
| 29 | Two pages maximum | One page costs you evidence. Three means you have not decided what matters |
| 30 | Tailor it to the ad, every time | Not a rewrite. Reorder the top third so it mirrors the language in the posting |
Long version of this stage: Resume Projects for Cloud and DevOps.
🔗 Stage 7. Your LinkedIn
| # | Step | Why it matters |
|---|---|---|
| 31 | Headline says the job you want, not the one you have | Within reason. "IT support, working towards cloud engineering" is honest and searchable |
| 32 | Post what you are learning, weekly | Not opinions, not motivation. What you built, what broke, what you worked out. It compounds |
| 33 | Comment more than you post | Comments on other people's posts reach more of the right people than your own when you are starting out |
| 34 | Connect with people doing the job, not only recruiters | The person two years ahead of you is more useful, and far more likely to reply |
| 35 | Turn on Open to Work | Visible to recruiters only, if you are worried about your current employer seeing it |
🔍 Stage 8. Finding the jobs
| # | Step | Why it matters |
|---|---|---|
| 36 | Apply before you feel ready | Feeling ready arrives about a year after being ready. Every ad is a wish list, not a specification |
| 37 | Twenty applications a week, not two hundred | Twenty considered beats two hundred sprayed. The spray approach gives you a number to quote and no interviews |
| 38 | Smaller companies hire juniors, big ones hire graduates | If you are a career changer, the graduate schemes were never your route in |
| 39 | Managed service providers are the fastest way in | They hire constantly, expose you to many client environments, and two years there is worth five somewhere quiet |
| 40 | Look for support roles that touch cloud | "Cloud support engineer" and "infrastructure analyst" are real jobs with real exposure and far less competition |
💬 Stage 9. The interview
| # | Step | Why it matters |
|---|---|---|
| 41 | They score how you narrow down, not what you know | Nobody expects a junior to have the answer. They expect you to have an order |
| 42 | Say what you would check first, and why first | "Effective routes, then firewall rules, then DNS" passes even when the cause turns out to be something else |
| 43 | Bring one story about something you broke | What happened, what you checked, what you changed, what you would do differently. Ready before you walk in |
| 44 | Ask what the environment actually looks like | How many subscriptions, what is automated, who gets paged. Tells you whether the job is real |
| 45 | "I do not know, here is how I would find out" is a passing answer | Guessing confidently is the failing one. Interviewers can tell immediately |
Labs for this stage. Troubleshooting, because the interview is a troubleshooting exercise wearing a suit.
Cloudlearn.ioConfigure monitor alerts and action groupsFinding out before the user calls you.labs.cloudlearn.io
Cloudlearn.ioApplication Insights logs and metricsReading what an application is actually doing.labs.cloudlearn.io
Cloudlearn.ioIntroduction to Kusto Query Language in Log AnalyticsQuerying logs. The skill that makes you useful in an incident.labs.cloudlearn.io
Cloudlearn.ioDocker container networking, bridges and DNSWhy the container cannot reach the thing. Classic interview ground.labs.cloudlearn.io
Cloudlearn.ioConfigure custom DNS on a virtual networkName resolution, and the failures that look like everything else.labs.cloudlearn.io
🏁 Stage 10. The offer, and after
| # | Step | Why it matters |
|---|---|---|
| 46 | Take the smaller title if the systems are real | Your title stops mattering in two years. What you touched does not |
| 47 | Ask about on-call before you sign | Frequency, compensation, and what actually pages. Not a dealbreaker, but never a surprise |
| 48 | First 90 days: read, do not change | The thing that looks obviously wrong is usually load bearing |
| 49 | Take the work nobody else wants | The report nobody automated, the thing someone set up and left. Fastest route to owning something |
| 50 | Write down what you learn, you will teach it later | Everyone I have watched get somewhere ended up explaining it to someone behind them. Start the notes now |
Labs for this stage. What you get asked to do in month one, which is almost always visibility rather than building.
Cloudlearn.ioConfigure diagnostic settings on Azure resourcesTurning on the logs nobody turned on. Common first-week task.labs.cloudlearn.io
Cloudlearn.ioApplication Insights for App Service monitoringVisibility into a running app.labs.cloudlearn.io
Cloudlearn.ioAzure Automation account basicsAutomating the repetitive thing you were hired to do manually.labs.cloudlearn.io
Cloudlearn.ioDeploy and manage Azure Key VaultWhere secrets belong, instead of a config file.labs.cloudlearn.io
⚠️ Where this goes wrong
The five failure modes behind almost every stall. Find yours, then jump back to the stage that fixes it.
| Symptom | What actually happened | Go to |
|---|---|---|
| Certificate, no interviews | Stage 4 done first, stage 5 never | Stage 5 |
| Nine half-finished projects | Started something new at every difficult part | Steps 21-22 |
| Cloud still feels impossible | Skipped networking and identity | Steps 6-7 |
| Applications, no replies | Resume lists technologies instead of outcomes | Steps 26-27 |
| Interviews, no offers | Answering with facts instead of an order of investigation | Steps 41-42 |
Warning
If you have been studying for more than a year without applying for anything, stop studying this week and apply for five jobs instead. The feedback loop from applying is faster than the one from another course, and you are almost certainly further along than you think.
Five years from helpdesk to first cloud job in my case. Some do it in two, some take seven. What separated the people who got there from the people who stalled was never intelligence and never starting young. It was finishing things.
Come back to this page and tick things off. It is meant to be worked through, not read once.
