Building large-scale Internal Developer Platforms: Core Principles
summary
An internal developer platform used by over 30 heterogeneous data teams centralizes GCP/AWS/Azure, GKE/Kubernetes, Prometheus/Grafana/OpsGenie, Argo CD, GitLab CI, Airflow, Terraform and Vault, and grows to ~400 GCP projects and ~100 GKE clusters in 2–3 years. It enforces dedicated cloud accounts (3–5 per team) for cost attribution and small blast radius, treats the offering as a product with roadmaps, support and user interviews, and stores inventory as a single source of truth driving automated terraform and onboarding.
AI-written summary · about
chapters (12)
- 0:00Introduction and context
- 1:09Platforms: Concepts vs technology
- 5:29Origin of software development at RTL
- 9:22Need for internal developer platforms
- 12:53First customer reactions and challenges
- 15:00Core principle: Treat platform as product
- 20:18Autonomy and dedicated cloud accounts
- 25:06Centralized inventory with CDB
- 28:20Balancing standardization and flexibility
- 31:57Building community and product management
- 34:03Key takeaway: Platform as product
- 34:24Q&A and closing
related talks
49m1 platform and multiple clusters, teams & countries. The Allianz case
15mEstablishing Platform Teams to Utilize the Economy of Scale
28m"Building the Plane While It's Flying"
41mBuilding the Future Trends in Modern Application Architecture
33mPaving the Road to Effective Software Development
27mDeveloper Platform Consoles Should Be Dumb
47mThe next frontier: Driving innovation with platform standardization
30mOrganizing Teams for GitOps and Cloud Native Deployments