Practical, hands-on lessons for cutting Kubernetes spend — rightsizing, autoscaling, and idle-resource elimination. Real labs, scripts, and AI-assisted workflows you can run on your own cluster this week.

A hands-on intermediate lesson covering fmt → validate → tflint → checkov → plan. Write constrained prompts that produce zero checkov findings, build a fail-closed validation script, wire it into pre-commit and CI, and add a cost gate with Infracost. 2.5-3.5 hours of labs.

Build an AI agent using boto3 + Claude API + Slack webhooks to detect AWS cost spikes before your bill lands. Intermediate lesson with full working code, 5 hands-on exercises, and a 12-item completion checklist.

Debug pod crashes, OOMKills, and CrashLoopBackOffs using LLMs. Build an AI diagnostic pipeline with kubectl + Claude, auto-triage your cluster, and cut mean-time-to-diagnosis from 15 minutes to under 2 minutes. Hands-on lab with 5 exercises, working scripts, and an agentic K8s debugger.

Build a production-ready Python AI agent that classifies Terraform state drift using Claude API — from terraform plan -json to auto-opened GitHub PRs. Includes full lab code, 5 exercises, and a Slack integration.

Build a 5-stage AI agent that catches AWS privilege escalation mid-chain — before the attacker finishes. Covers 6 real escalation patterns (LAMBDA_PASSROLE, POLICY_VERSION, TRUST_HIJACK and more), Shannon entropy pre-filter, LLM triage with structured JSON output, and GitHub Actions deployment with OIDC auth. Intermediate-Advanced. 3-4 hours.

Build a complete production system: boto3 + AWS Cost Explorer + Cost Anomaly Detection + Claude API to detect, explain, and Slack-alert cost spikes. Intermediate, ~4 hours. Includes 6 hands-on exercises, a 452-line production Python script, Slack Block Kit delivery, Lambda scheduling with Secrets Manager, and a 5-day homework. Members typically find $500-$3,000/month of recoverable waste in their first week.

Build a complete VPC + EKS + RDS stack using Claude AI as your Terraform co-pilot. Intermediate, ~3 hrs. 6 exercises, real code, pre-commit review bot.

Build a complete pipeline that takes a CloudWatch, Datadog, or PagerDuty alert payload and generates a structured incident runbook via Claude API in under 60 seconds. Full lab with Python scripts, prompts, and CI integration.

Cut your logging and metrics bill 40-60% with an AI-assisted cardinality review. Measure Prometheus cardinality via the TSDB status API, break down CloudWatch and Cloud Logging ingest per log group, triage hundreds of streams with a hard-constrained LLM prompt, then ship reductions safely with metric_relabel_configs, OTel processors and exclusion filters. Includes the never-drop guardrail for audit and compliance logs, and a CI gate that blocks new unbounded labels.

Build a token-metering LLM gateway on Kubernetes with per-team virtual keys, Prometheus spend metrics, and automated budget guardrails that degrade or block runaway agent workloads. Full lab with runnable Terraform, Helm, Python hooks and PromQL.

Turn a free RVTools export into a defensible VMware migration business case. Broadcom's per-core subscription and the 72-core order-line minimum reshaped the maths. This lab shows you how to count what you actually run, price it, and decide. Full Python analysis script, the licensing arithmetic, a 3-year TCO model across GCVE, AVS, Nutanix, Proxmox and native refactor, plus the AI prompts for the wave plan. Runs on your laptop against your own estate. $0 cloud spend.