The Challenge
Luigi’s Box builds AI-powered search, recommendations and product listing for e-commerce, used by online stores across Central Europe and beyond. Search sits on the critical path of every store that uses it, so the platform behind it has to be reliable, observable and easy to change safely.
From June to September 2026 I worked as an Infrastructure Consultant in their core infrastructure team on a three-month contract. I joined an existing, already configured Kubernetes setup, so the focus was on operating it well and shipping changes safely rather than building it from scratch.
The Approach
The work centred on the production Kubernetes clusters and everything around them:
- Kubernetes operations: running and maintaining production clusters, deploying and configuring internal services so they behaved the same way across environments.
- GitOps delivery: every change went through Git. GitHub Actions built and tested, ArgoCD synced the cluster to the desired state, so deployments were automated, reviewed and auditable.
- Infrastructure as Code: AWS resources provisioned and changed through Terraform instead of the console.
- Observability: Grafana dashboards for infrastructure and internal services, so problems show up on a graph before they show up in a support ticket.
Technical Highlights
I built an internal GitHub overview service in Python that collects engineering metrics from GitHub and exposes them to Grafana. Delivery data sat next to infrastructure metrics on the same dashboards, without anyone exporting spreadsheets.
The role built on my earlier infrastructure work, such as moving Exlink from Windows/IIS servers to Linux, then to Hetzner and onto a Kubernetes cluster I built from scratch, and applied it at SaaS scale on AWS.
Results
- Production Kubernetes clusters operated as part of the core infrastructure team
- GitOps workflow with GitHub Actions and ArgoCD for automated, auditable deployments
- AWS infrastructure managed as code with Terraform
- Engineering metrics from GitHub made available in Grafana through a custom Python service
