Roadmap 2027 · 12 checkpoints
DevOps & cloud developer roadmap 2027
You like systems, automation and keeping things running. The path: Linux → Docker → CI/CD → AWS → Terraform → Kubernetes. Every checkpoint has topics, a project, how to use AI, and a test that tells you when to move on.
Stage 1 · Foundations
- 4 wks
Linux & the shell
Every server, container and CI runner is Linux. This is the floor everything stands on.
Files & permissionsProcesses & signalssystemdPackage managersBash scriptinggrep, awk, sedSSH & keys- Build
- A Bash script that backs up a folder, rotates old backups and logs what it did, run by a systemd timer.
- Use AI
- Ask AI to explain a command flag by flag before you run it on a real machine.
- Cleared when
- You can find why a service is down using only the shell and its logs.
- 3 wks
Networking for DevOps
Most outages are DNS, TLS or a firewall. Know the layers well enough to debug them.
TCP/IP & portsDNS recordsHTTP & TLS certificatesLoad balancersReverse proxies (Nginx, Caddy)Firewalls & security groups- Build
- Serve two apps on one VPS behind a reverse proxy with automatic HTTPS.
- Use AI
- Paste a curl -v or dig output and ask AI to walk you through each line.
- Cleared when
- You can explain why a site is unreachable using dig, curl and ss.
- 2 wks
Git, a scripting language & YAML
Automation is code. Pick Python (or Go) for tooling and get fluent with YAML.
Git branching & PRsPython for scriptsCalling REST APIsYAML & JSONRegular expressions- Build
- A Python script that checks a list of URLs and posts failures to a Slack or Discord webhook.
- Use AI
- Write the script yourself, then ask AI to review error handling and timeouts.
- Cleared when
- Your script handles timeouts, bad responses and retries cleanly.
- 3 wks
Containers with Docker
Containers are how almost all software ships in 2027.
Images & layersDockerfile best practicesMulti-stage buildsVolumes & networksDocker ComposeImage scanning- Build
- Containerise a web app and its database with Compose, with a small, non-root production image.
- Use AI
- Ask AI why your image is large, then measure each change with docker image ls.
- Cleared when
- Your production image is small, runs as non-root and starts with one command.
Stage 2 · Build & ship
- 3 wks
CI/CD
Shipping safely many times a day is the core of the job.
GitHub ActionsBuild, test, scan, deploy stagesCachingSecretsEnvironments & approvalsRollbacks- Build
- A pipeline that tests, builds an image, pushes it to a registry and deploys to staging, then to production on approval.
- Use AI
- Ask AI to review your workflow for leaked secrets and missing steps; check each claim.
- Cleared when
- A merge to main reaches production without anyone touching a server.
- 5 wks
Cloud fundamentals (AWS first)
Most DevOps jobs in India ask for AWS. Learn one cloud properly; the others map across.
IAM & least privilegeVPC, subnets, routingEC2 & autoscalingS3RDSLoad balancersBilling alarms- Build
- Deploy your app on AWS in a private subnet behind a load balancer, with a billing alarm set up first.
- Use AI
- Ask AI to explain an IAM policy line by line before you attach it.
- Cleared when
- Nothing has rights it does not need, and you know what you are paying for.
- 4 wks
Infrastructure as Code
Clicking in a console does not scale or get reviewed. Infrastructure belongs in Git.
Terraform or OpenTofuState & remote backendsModulesPlan & apply in CIAnsible basicsDrift- Build
- Rebuild your AWS setup with Terraform or OpenTofu modules, applied from CI after a reviewed plan.
- Use AI
- Generate a first module with AI, then read the plan output line by line before applying.
- Cleared when
- You can destroy and recreate the whole environment from Git.
- 6 wks
Kubernetes
The standard way to run containers at scale. Interviews go deep here.
Pods, Deployments, ServicesConfigMaps & SecretsGateway API (Ingress NGINX was retired in March 2026)Probes & resourcesHelmAutoscalingkubectl debugging- Build
- Run your app on a local cluster (kind or k3d), then on a managed cluster such as EKS, with Helm.
- Use AI
- Ask AI to explain why a pod is in CrashLoopBackOff from its describe and logs output.
- Cleared when
- You can deploy, scale, roll back and debug a broken pod without a tutorial.
Stage 3 · Get hired
- 2 wks
GitOps & safe releases
Teams want every change to the cluster to come from Git, with safe rollouts.
Argo CD or FluxEnvironments as foldersBlue/green & canaryFeature flagsRollback drills- Build
- Manage your cluster with Argo CD and run a canary release.
- Use AI
- Ask AI to compare canary and blue/green for your app, then pick one yourself.
- Cleared when
- Reverting a commit rolls production back.
- 3 wks
Observability & on-call
You are judged by how fast you find and fix problems.
Prometheus metricsGrafana dashboardsLogs (Loki or ELK)OpenTelemetry tracesAlerts & SLOsPostmortems- Build
- Dashboards and alerts for your app; then break it on purpose and write a postmortem.
- Use AI
- Ask AI to help write a PromQL query, then check it against the raw metrics.
- Cleared when
- An alert fires before users notice, and the dashboard shows why.
- 2 wks
Security & DevSecOps
Supply-chain attacks and leaked keys are among the most common incidents.
Secrets managersImage & dependency scanningSBOMsPolicy as codeLeast privilege in CIBackups & restore tests- Build
- Add scanning, signed images and a secrets manager to your pipeline; test a restore from backup.
- Use AI
- Never paste real keys into a chat. Ask AI how a leaked key would be abused, then rotate yours.
- Cleared when
- No secret lives in Git, and a restore from backup actually works.
- 3 wks
AI platforms & interviews
Teams now run AI services too. Platform engineers own their cost, scaling and safety.
Serving model APIsGPU basics & costRate limits & cachingPlatform engineeringSystem design for infraIncident scenarios- Build
- Put an LLM-backed service behind your gateway with rate limits, caching and a cost dashboard; write 3 infra designs.
- Use AI
- Use AI as an interviewer for incident scenarios. Answer first, then compare.
- Cleared when
- You can design a highly available setup for a web app and explain what it costs.
Learning with AI
Do this, not that
Do
- Try for 20 minutes before you ask AI.
- Ask "why" until you could teach it back.
- Read every diff an assistant makes, then run the tests.
- Rebuild one AI-written feature from scratch each month.
- Practise DSA and machine coding with AI off.
Don't
- Pasting API keys, passwords or company code into a chat.
- Shipping code you can't explain line by line.
- Letting AI pick your architecture before you know the options.
- Trusting version numbers or APIs it states without checking the docs.
- Using AI in online assessments unless it is allowed.
DevOps & cloud roadmap questions
How long does the devops & cloud roadmap take?
About 9 months at 2.5 hours a day, 6 months at 4 hours a day, or 15 months at 1.5 hours a day. It has 12 checkpoints; move on when you pass each "cleared when" test, not when the weeks run out.
What will I build on the devops & cloud roadmap?
One project per checkpoint, 12 in total, from "A Bash script that backs up a folder, rotates old backups and logs what it did, run by a systemd timer." to "Put an LLM-backed service behind your gateway with rate limits, caching and a cost dashboard; write 3 infra designs."
When should I start applying for jobs?
Often, yes. Most people start applying after the "Build & ship" stage, once they have two or three real projects. The "Get hired" stage (system design, interviews, a capstone) is what moves you from a first offer to a better one.
Should I learn with AI tools or without them?
Both, in that order. Use AI as a tutor from day one, as a pair programmer once you build your first project, and as a debugger always. Keep DSA, machine-coding practice and interviews AI-free, and never ship code you cannot explain.
Do I need DSA for these roadmaps?
For frontend, backend, Java and full stack, yes: solve one problem a day from the DSA sheet alongside every checkpoint, because most product companies in India still have a coding round. DevOps and game studios ask less DSA, but basic arrays, hashing and complexity still come up.
I am from a non-CS branch or a tier-3 college. Will these work for me?
Yes. Every track starts from the foundations and is judged by what you build, not your degree. Put your projects on GitHub, deploy them, and write short READMEs: that is what recruiters and interviewers look at first.
Which versions does the 2027 roadmap use?
React 19, Next.js 16, Node.js 24 LTS, Java 25 LTS and Spring Boot 4 for the web tracks; Unity 6.3 LTS or Godot 4.6 for games; and current Kubernetes with the Gateway API for DevOps, since Ingress NGINX was retired in March 2026. Check the official docs before you start, because versions move.