The Platform
Cloud infrastructure, Kubernetes platforms, delivery systems, observability and secrets — from provisioning the environment to carrying the pager for what runs in it.
From source to production
The stack as I have run it — edge, cloud, orchestration, services, state and the control plane holding it together. Select any layer to open its architecture.
The stack, edge to state
Select a layer for its architecture
Edge
Cloud
Orchestration
Services
State
Control
The platform, not the command. Workload shape, traffic path, storage, identity and limits — the parts that decide whether a cluster survives its second year.
Ingress
TLS termination and host routing
Service
Stable address in front of moving pods
Deployment
Rollout strategy, probes, resource limits
Pods
Requests sized from observed usage, not guesses
Config / secrets
Mounted, never baked into the image
Persistence
Claims, storage classes, backup path
Worked with
Delivery, Kubernetes, secrets, observability and what happens when it breaks — the same system seen from five angles. Select one to open it.
One path to production, identical in every environment. Scanning is a gate rather than a report, and the artefact that passed the tests is the one that ships.
The pipeline
Git push
Pipeline
GitLab CI / GitHub Actions
Test
Security scan
SAST · secret scanning · Helm lint
Docker build
Tagged by commit
Registry
ECR
Deploy
Helm · approval-gated
Health checks
Probes before traffic
Observability
Traces confirm the change
What I hold it to
Worked with
Reviewed as a diff, applied from CI, and reconciled back to the file when it drifts — the same lifecycle whether the target is a cloud account or a rack with no internet.
The lifecycle
01
Terraform
Desired state in review
02
Plan
The diff, read before it runs
03
Review
A pull request, not a console click
04
Apply
From CI, with state locked
05
Drift
Detected, not discovered
06
Reconcile
Back to the file, never by hand
Worked with
Infrastructure is code, not a sequence of console clicks. If an environment cannot be rebuilt from its source, it isn’t finished — which is why one reviewed script stands the whole thing up, whether it lands on a cloud account, bare metal or an air-gapped rack.
Not production. Not a demo reel.Small systems I build to test ideas.
01
When a system fails
Failure → telemetry → mitigation → recovery
02
When infrastructure drifts
Desired state → drift → plan → reconcile
03
From requirements to architecture
Requirements → architecture → validation
Small, working models of the problems I spend my time thinking about.
Enter the lab