A small firm that measures what it operates.
Kernforce runs and improves cloud platforms — Kubernetes, CI/CD, AWS infrastructure — for teams that already have them. We are deliberately small, we say what we cannot do, and we build the tooling we use to measure our own work.
How we work
Four things we will not trade away
Measure first, and keep measuring
We do not estimate environments we cannot observe. Every engagement starts with measurement, every number we publish is traceable to something you can check yourself, and where the data does not exist we say so rather than filling the gap with an assumption.
Propose, do not seize
Read-only access by default. Changes arrive as something you review and approve — a pull request, a plan, a scheduled window — not as a controller with admin rights making decisions on your behalf at three in the morning.
Narrow the claim
It is easy to produce a large savings figure and hard to defend one. We reconcile findings against each other, discard advice that the reader cannot execute, and hand over the number that survives review — which is usually a good deal smaller, and worth more because it gets acted on.
Leave the knowledge behind
Infrastructure as code in your repository, runbooks written to be executed by someone who did not write them, and decisions recorded with the alternatives that were rejected. If an engagement ends and your team cannot carry on without us, we did it wrong.
Our tooling
We build our own tools
We develop MetrixForge, a platform for measuring Kubernetes cost, efficiency and risk: agents that collect metrics and scan images, an analytics layer over the result, and recommendations that can be applied as a reviewed pull request against your own repository.
It exists because we needed it. When we tell a client what something costs, we can show how the figure was computed — and when a recommendation is wrong, we can fix the thing that produced it instead of apologising for it.
MetrixForgeStack
What we work with
The list is what we have run in anger, not everything that exists.
- AWS
- Amazon EKS
- GKE
- Hetzner
- Kubernetes
- Karpenter
- Cluster Autoscaler
- Helm
- Cilium
- Terraform
- OpenTofu
- Argo CD
- Flux
- GitHub Actions
- GitLab CI
- Prometheus
- Grafana
- OpenTelemetry
- VictoriaMetrics
- Velero
AWS
How we work on AWS: account and network design, cost attribution that reconciles to the bill, and migrations sized from measured demand.
Kubernetes
How we work with Kubernetes: requests as the unit that matters, reliability settings that only count on the day they count, and measurement before change.
Karpenter
Karpenter in production: NodePool design, consolidation behaviour, disruption budgets, and why node-level cost advice usually points at the wrong lever.
Small enough to answer the phone.
A team of senior engineers, and the first call is a technical one. The work is written down as it happens, so it carries on if one of us is ill or leaves.