DevOps / Cloud Platform ready

An outcome-driven course for working engineers moving into a DevOps or Cloud Platform role — but with a specific modern angle: you don't just learn the domain, you learn it through Claude Code and leave with a library of seven reusable Claude skills you built and validated yourself. Seven modules, each ending in a hands-on task you drive through Claude on your own laptop. The payoff is cutting time-to-response and time-to-action when it matters: when a page comes in, you're reaching for a skill you already built and validated, not starting from a blank terminal. No theory you don't immediately apply, and no copy-pasting without reading.

By the end of the course you can, driving through Claude Code and understanding every file it writes:

1. Deploy a containerised service to a Kubernetes cluster using kind locally.
2. Package that service as a Helm chart with configurable values.
3. Wire a GitHub Actions pipeline that builds the image, runs tests, pushes to a registry, and deploys.
4. Provision a cluster with Terraform and know how state drift bites you in real work.
5. Instrument observability — Prometheus metrics + a Grafana dashboard + a scrape-gap alert.
6. Debug the three canonical pod failures (CrashLoopBackOff, ImagePullBackOff, OOMKilled) and know the shape of a healthy pod.
7. Scope a ServiceAccount with least-privilege RBAC and sketch Workload Identity Federation for cloud IAM.

Each module has four lessons: a short concept explainer, a Claude-guided hands-on task with an acceptance test, a skill-building lesson where you codify the session as a reusable .claude/skills/<name>/ skill and validate it in a fresh Claude session against both valid and adversarial inputs, and a recall quiz. Expect ~9 hours of focused work across the course.

How this course works (the three principles you're signing up for)

• Stay close to your problem. Claude types fast; that is never a license to disengage. Every task lesson requires you to read every file Claude produces before you run it, and to ask Claude to explain anything you don't understand. The skill you build at the end of each module is valuable only because you understood the domain underneath it.
• Validate what Claude will NOT do, not just what it will. Every skill you produce ends with two sentences: "This skill handles: …" and "This skill does NOT handle: …". A skill that documents only its capabilities is dangerous in the hands of users who don't yet know its limits. You'll test every skill in a fresh Claude session against deliberately wrong inputs to confirm it fails loudly rather than silently doing the wrong thing.
• Design for realistic work environments. Pristine demos lie. Every module includes a deliberate "break it on purpose" probe — a targetPort mismatch, a missing CI permission, a state-drifted cluster, a silently-broken ServiceMonitor — so the failure modes your skill documents come from lived experience, not guesses. The realistic moment we're designing for is the one with a clock on it: an alert fires, and you need to move from what's broken? to what's the fix? without starting from zero.

What you need before you start

• A laptop with Docker Desktop (or Podman) installed.
• Claude Code installed and authenticated — every lesson assumes you can open a Claude Code session in a working directory.
• kubectl and kind on your PATH. Install steps are in module 1.
• A GitHub account (used from module 3 onward).
• A free-tier cloud account with one provider (GCP, AWS, or Azure) for module 4's bonus + module 7's federation sketch.
• Comfort reading YAML and running terminal commands. Comfort reading what Claude wrote before running it.

What this course is not

• Not a Kubernetes reference — the point is to build the reflexes, not memorise every resource type.
• Not a prompt-engineering course — the focus is producing reliable skills that capture your process, not writing cleverer prompts.
• Not a theory course — you won't get a full explanation of consensus, CNI internals, or the control plane unless a task needs it.
• Not a "let Claude do the work" course. Every lesson insists you read, understand, and validate what Claude produces. If you're not willing to engage that way, this isn't the right course.
• Not assessed by a certification — it's assessed by whether you can do the seven tasks on demand and hand over seven working skills that survive a fresh Claude session.