[ Google Cloud Platform ]
Building your Kubernetes-native platforms on GCP
Kubernetes platform
Leveraging cloud-native Kubernetes with GKE
Kubernetes was developed by Google, and GKE remains the most mature managed solution on the market. We operate your clusters in Autopilot or Standard mode, for data platforms, B2B SaaS solutions and high-traffic back-office systems. GKE is not Kubernetes-compatible: Kubernetes is GKE-compatible.
Architecture as Code
Standardising your environments using Terraform and workspaces
We model your Dev, Pre-production and Production environments using Terraform for workspace management, ArgoCD for GitOps, and native integration with managed Cloud SQL and Cloud Storage services. A replicable platform is not a platform: it’s a product.
GCP FinOps
Anticipating the potential cost pitfalls specific to GCP
Egress, IOPS/storage pairing, Committed Use Discounts: there are particular pitfalls it is easy to fall into with GCP costs. We design FinOps-compliant architectures right from the start. Cost transparency in GCP is not automatic: it has to be intentionally built.
The cornerstones of our GCP expertise
- GKE Autopilot and Standard
GKE Autopilot offers the benefits of managed Kubernetes without the day-to-day operational burden: pod-based improvements, GPU and TPU support for specific workloads, and per-pod-request pricing. We use whichever mode best meets your requirements: stateful workloads in Standard mode, and scalable workloads in Autopilot mode.
- Terraform with workspace management
Our preferred pattern on GCP is a single Terraform repository, workspaces for each environment, shared modules, and CI/CD pipelines that apply changes after review. The result: perfectly consistent and reproducible environments.
- Apigee and API governance
Apigee is the leading API management solution on GCP, following Google’s acquisition of the platform. We pride ourselves on our expertise in large-scale API governance, including analytics, monetisation and granular security policies, alongside Kubernetes and Cloud Run services.
- Cloud Load Balancing and Cloud Armor
Global Cloud Load Balancing and Cloud Armor (rate limiting, Adaptive Protection) form the native GCP security stack. We block Layer 7 attacks without compromising on application latency or generating crippling false positives.
- Cloud SQL and managed services
Cloud SQL for MySQL and PostgreSQL, Cloud Filestore for NFS requirements, Artifact Registry, Secret Manager. The right building blocks managed in the right place, without making the architecture dependent on proprietary primitives.
- Observability, Cloud Logging and Loki on GCS
We combine native GCP services (Cloud Logging, Cloud Monitoring, Cloud Trace) with Grafana, Prometheus, and Loki on GCS and OpenTelemetry to provide comprehensive observability. For multi-cloud requirements, Thanos or Mimir consolidate the data sources.
- AI in delivery practice
Apart from Vertex AI, we use Claude Code, Gemini CLI and MCP on a daily basis to speed up migrations, upgrades and incident investigations. For us, it’s another tool in the stack, not a selling point.
[‘’]
GCP is the only one of the three major hyperscalers to openly acknowledge the role Kubernetes plays. It makes sense: Kubernetes was developed at Google. When you inherit a Cloud SQL and GKE platform, the operational experience is more straightforward than elsewhere: fewer proprietary abstraction layers, more pure Kubernetes patterns.
The downsides are the billing pitfalls related to egress and the coupling between IOPS and storage, which must be anticipated from the architectural design stage. This is where we come in: mastering the technical patterns needed to capitalise on GCP’s true Kubernetes-native superiority, without letting the bill spiral out of control unnoticed.
Adrien Bresson, Head of Cloud Infrastructure
With GCP, cost transparency is not automatic. It has to be built in right from the design stage, just like portability.
