Nokia
Working Student DevOps Engineer · Jul 2024 - Sep 2026 · Poland
AWS infrastructure automation - VPC, EKS, RDS, Crossplane, Terraform
- Problem
- Infrastructure was provisioned as a monolithic Terraform stack, coupling VPC, EKS, RDS, CloudFront, ElastiCache and application deployment without reusable modules, making multiple environments harder to reproduce and manage.
- What I did
- Built a modular, Terraform-provisioned control-plane EKS cluster with built-in AWS add-ons and pre-installed Crossplane. Created reusable XRDs and Compositions to provision VPC, EKS, RDS, CloudFront, and ElastiCache resources as an umbrella deployment for production and development environments.
- Result
- Infrastructure became reproducible, declarative, and continuously reconciled. Multiple environments could be created using the same XRD/Composition patterns, reducing setup time to ~15 minutes.
GitOps Delivery - Argo CD, Gitea, Crossplane, Helm
- Problem
- Infrastructure, third-party services, and application changes lacked a fully auditable, Git-driven delivery process with continuous state reconciliation.
- What I did
- Implemented GitOps workflows using Argo CD, internal Gitea, and Crossplane, with infrastructure and application state defined declaratively in Git and deployed through Helm.
- Result
- Established auditable, Git-driven delivery with continuous reconciliation for infrastructure, third-party services, and application environments.
Observability - Prometheus, Grafana, CloudWatch
- Problem
- EKS monitoring relied on a licensed Datadog solution, while development environments needed a self-hosted monitoring stack to reduce costs and integrate monitoring with existing deployments.
- What I did
- Implemented Prometheus and Grafana with ALB ingress, EBS-backed persistent storage, application-specific ServiceMonitors and dashboards, and CloudWatch integration via IRSA for detailed RDS and EKS node monitoring. Packaged the configuration into reusable Helm charts.
- Result
- Delivered a production-ready, license-free monitoring stack automatically deployed with Helm, significantly reducing monitoring setup effort and operational cost.
Cost Optimization - Cluster Autoscaler, shared ALB
- Problem
- Development EKS nodes were running continuously outside business hours, cluster scaling was performed manually, and the environment used nine separate ALBs for ingress.
- What I did
- Automated EKS node scaling using Cluster Autoscaler, Pod Identity, and Metrics Server, with scaling rules configured to reduce development capacity to zero outside business hours. Consolidated ingress target groups from nine ALBs into one shared ALB using the ALB group configuration.
- Result
- Enabled automated on-demand cluster scaling, eliminated unnecessary off-hours compute usage, and reduced ALB costs by ~90% (9 → 1).
CI/CD - Jenkins, Helm
- Problem
- Jenkins pipelines and Helm deployment workflows required continuous maintenance as infrastructure and application components evolved.
- What I did
- Created, maintained, and enhanced Jenkins pipelines, troubleshot deployment issues, and updated Helm-based deployment workflows to remain aligned with infrastructure and application changes.
- Result
- Kept CI/CD workflows reliable and synchronized with ongoing infrastructure and product changes across environments.