I build production-style infrastructure — the part that autoscales, self-heals, and keeps the pager silent. AWS, Kubernetes, and Terraform, wired with real operational discipline.
“Build production-grade systems. Sleep through the on-call.”
Tap a card to expand into the skills the exam actually tested.
Each entry links to a full case study — architecture, CI/CD, infra, and what I learned when it broke.
CKA-certified · B.Tech CSE student · open to DevOps and Cloud Engineering internships. Email lands fastest.
kubectl apply on a tutorial.Here’s what I did differently — with the receipts to back it up.
I did not copy-paste a blog post. I enabled the OIDC provider, wrote the trust policy with the service-account condition, scoped the IAM policy to a specific table ARN, annotated the ServiceAccount, and verified it by exec-ing into the pod and checking env | grep AWS. I know it works because I know why it works.
I wrote a module that other modules consume. Input validation, dual deployment modes, region-agnostic. When I call it from RoboShop and again from a new project, I don't copy-paste HCL — I reference the same tested module.
I did not write a Jenkinsfile per service. I wrote three reusable Groovy functions that handle version detection for Maven, Go, and Node — then one pipeline definition that every service calls. When the deployment process changes, I change it once.
I did not add them because a tutorial said to. I added them, measured the deploy time, watched it drop from 12 minutes to 1.5 minutes, and understood why. The startup probe is what made the difference — Kubernetes was killing pods before they finished connecting to databases.
From bare Linux to production-style Kubernetes — a path built on purpose, not shortcuts.
RoboShop on EC2 + ALB. 15 Terraform modules, ASGs, security groups, ansible-pull. No shortcuts — provisioned every subnet, wrote every SG rule.
Terraform VPC module — dual testing/prod mode, region-agnostic, input-validated. Consumed by every downstream project instead of copy-pasted.
Kubernetes lab from first principles — every resource type, every concept, hands-on manifests. Pods → Deployments → Services → Ingress → RBAC → NetworkPolicy.
Retail Store on AWS EKS. Helm umbrella charts, IRSA with a scoped IAM policy, HPA at 70% CPU, StatefulSets with EBS dynamic provisioning, full observability.
Jenkins shared library — one pipeline definition, three Groovy functions, every microservice in every language. Deploy process changes in one place.
Passed the Certified Kubernetes Administrator exam. Bootstrapping, upgrades, RBAC, etcd backup/restore, and troubleshooting from first principles.