Cloud SQL was doing two very different jobs: application CRUD and analytical pipeline data. Moving the analytical side to BigQuery — and keeping Cloud SQL where it belongs — gave each workload a better home and a better cost model.
An over-provisioned GKE cluster, large fixed nodes sitting mostly idle — and how Node Auto-Provisioning plus honest resource requests cut the cost by roughly 50% without redesigning the platform.
From React and React Native to data pipelines, GCP, GKE, GitLab CI/CD and platform engineering — what changed when I stopped looking only at the screen and started seeing the whole system behind it.