Kubernetes & Container Platforms
AKS and EKS platforms: networking, ingress, identity, autoscaling, upgrades and security, kept repeatable with GitOps.
The problem
A cluster is easy to create and hard to run well. Teams end up with hand-tuned clusters nobody dares upgrade, network policy that grew by exception, secrets in the wrong place, and no way to build a second cluster or region the same way.
What you receive
- Platform reference architecture and cluster standards
- Cluster build defined in Terraform and Helm/Argo CD, reproducible per environment
- Operational runbooks: upgrades, scaling, incident response
- Hardening and cost findings with a ranked remediation list
Capabilities
- AKS and EKS architecture and cluster lifecycle strategy
- Multi-cluster and multi-region platform design
- Ingress, service mesh where justified, network policy and egress control
- Storage, scheduling, resource requests and autoscaling (HPA, KEDA, cluster autoscaler, Karpenter)
- Secrets management and workload identity (no long-lived credentials in pods)
- RBAC, admission policy and cluster hardening
- Observability for the platform and the workloads on it
- Upgrade planning and rehearsed, boring upgrades
- Workload migration onto the platform
- Progressive delivery and internal developer platform foundations
Example engagement
Platform rebuild for a SaaS company with eleven snowflake clusters
Defined a single cluster standard, rebuilt it as a Terraform module with Argo CD bootstrapping the platform add-ons, migrated workloads cluster by cluster, and left the team with an upgrade runbook they have since used unassisted.
Representative of the work; client details are not published.
Technologies
- AKS
- EKS
- Helm
- Argo CD
- Flux
- KEDA
- Karpenter
- Cilium
- cert-manager
- External Secrets
- Kyverno
- Gatekeeper
Related solutions
Reduce cloud spend
Find what drives the bill across architecture, Kubernetes, storage, databases, builds and observability, then change it.
Build a Kubernetes platform
An enterprise Kubernetes environment with networking, security, GitOps, observability, autoscaling and operating standards.
Harden the software supply chain
Integrate security scanning, SBOMs, policy gates, signing, secrets detection and container controls.
Build an internal developer platform
Repeatable onboarding, deployment templates, service standards, infrastructure abstractions and developer self-service.
Multi-region and high-availability architecture
Improve availability across cloud regions and failure domains, with recovery you have rehearsed.
How we think about it
Kubernetes · Security · Platform engineering · 8 min
Isolating tenants on a shared Kubernetes cluster
A namespace is not a security boundary, but a namespace plus five other controls is a usable one. What separates tenants on a shared cluster, what still needs a cluster of its own, and how to decide which customer gets which.
Kubernetes · Cloud cost · 8 min
Kubernetes rightsizing: where the money actually goes
Most Kubernetes clusters are paid for twice: once for the nodes and once for the requests that keep them half empty. A practical order for finding and fixing it.
Before you write to us
The questions we get asked first.
Answered the way we would on a call. The consulting FAQ covers how engagements work in general.
- Shared cluster or a cluster per team?
- Both, with a rule for who gets which. A shared cluster with default-deny network policy, quotas, restricted pod security, no shared service accounts and separate data per tenant is the default; a dedicated cluster is for workloads that need cluster-level privileges, a specific region or contractual separation.
- Can you work with the cluster we already run?
- Yes. The first step is scoring the existing cluster against the isolation and operations controls, not replacing it. Upgrades, node-pool shape, autoscaling and the delivery path onto the cluster are usually where the work is.
- What access do you need?
- Read access to the cluster (a read-only kubeconfig), the manifests or Helm charts, and the pipeline that deploys to it, and a few hours of the team's time. Nothing is run under your production credentials from outside your own tooling, and write access only arrives with an implementation phase, through your pipeline and your review.
- Who does the work?
- Senior engineers who have run this kind of estate, working in your repositories, pipelines and rituals. There is no hand-off from a sales team to a delivery team: the people on the first call are the people on the engagement.