AI
From Early Adopters to a Scaling Problem
The Belastingdienst started building on Kubernetes in 2017 with one goal: maximum developer autonomy. By 2019, early adopter teams were landing on the platform and velocity was high. Then growth exposed a gap. Developers arriving after that phase had never touched OpenShift, Argo CD, or network policies. Onboarding that once took hours stretched to weeks or months. Enablement couldn’t keep pace with demand, and platform growth stalled. The team needed a way to cut cognitive load without reducing what the platform could do.
Project as a Service: One YAML, Full Environment
The answer was a Kubernetes operator called Project as a Service (PaaS). A DevOps team forks a repo, fills in one YAML file with namespace names, resource quotas, and the capabilities they want (Argo CD, Keycloak, Grafana), then opens a pull request. A validator checks the YAML structure, LDAP queries, and secret references. First-time requests require two approvers. Quota changes also need human sign-off to protect cluster capacity. Once merged, the operator provisions everything in roughly two minutes. The operator is fully open source on GitHub.
The Backstage Starter: CI/CD Pipelines Without Manual Wiring
Alongside PaaS, the team ships a Backstage template they call the starter. Running it produces five Git repos: a code repo, a test repo, a Tekton pipeline repo, a deploy manifests repo, and an Argo CD repo. The Tekton repo arrives pre-populated with tasks for SonarQube, Trivy, and the build step. The Argo CD repo contains application definitions ready to sync. The starter also creates the PaaS YAML and opens the pull request automatically. Developers get a working CI/CD chain without writing a single pipeline definition.
The Golden Path: 50-Plus White Papers Across the App Lifecycle
Tooling alone does not replace knowledge. The team maintains over 50 white papers organized across day zero (architecture and security decisions), day one (GitOps deployment and PaaS setup), and day two (scaling, right-sizing, runtime security, incident management). These are updated continuously. Alongside the white papers, the team runs a community of practice for lead developers, a container user group for release demos, self-paced Tekton workshops at basic and advanced levels, and a one-day immersive program where a team builds and deploys a real application on the platform from scratch.
Scale: 78 Clusters, 900 Nodes, Exponential Growth After November 2023
Platform growth was steady from 2021 through mid-2023, then accelerated sharply. The inflection point was November 2023, when the enablement team expanded and adoption became an explicit goal alongside infrastructure work. Today the platform runs 80-plus teams, 160 production applications, 78 OpenShift clusters, 227 development namespaces, 125 production namespaces, and over 900 worker nodes. Projections put the cluster count past 100 this quarter. A team of roughly 40 people operates the whole thing, several of whom were in the audience at the time of this talk.
Notable Quotes
our fully open-source solution to giving developers a full Kubernetes environment with just one YAML request. Jerry van Hulst · ▶ 09:01
Developers don’t need to install or configure anything. Just go to your pass, fill in. I want Argo CD, I want Graphana, maybe you want Keycloak. Just put it in there. Jerry van Hulst · ▶ 10:25
Developers don’t need 10 different ways to do something a little bit. They want one way to do something good. Jerry van Hulst · ▶ 24:14
I’m glad you brought up AI because it wouldn’t be a tech conference if you didn’t mention it at least once. Jerry van Hulst · ▶ 23:45
Key Takeaways
- One YAML pull request provisions a full OpenShift environment with Argo CD, Keycloak, and Grafana in under two minutes.
- Platform growth was flat until enablement investment in November 2023, then reached 78 clusters and 900 nodes.
- Self-sufficiency, not hand-holding, is the stated goal: 50-plus white papers and self-paced workshops replace repeated support tickets.
About the Speaker(s)
Jerry van Hulst is Product Owner of the Container Hosting Platform at the Dutch Tax Authority (Belastingdienst). He leads the team building and scaling the Kubernetes-based platform that now serves 80-plus internal DevOps teams.
Marcel Kerker is a CNCF Ambassador and DevOps Consultant at HCS Company B.V. He focuses on cloud-native technologies and has spoken at industry meetups and conferences across the cloud community.