PLAIN: A Sovereign Data and AI Platform for Germany
Five ministries, zero shared infrastructure
Germany's federal ministries were sitting on valuable data they couldn't use together. Each department ran its own tools, its own pipelines, its own contracts with vendors. Cross-ministry analysis on topics like pandemic response or climate policy meant months of coordination, if it happened at all. Meanwhile, sovereignty requirements ruled out the usual hyperscaler shortcuts.
The platform that didn't exist yet
PLAIN (Platform Analysis and Information Systems) needed to serve five federal ministries from a sovereign Berlin data centre, operated by Bundesdruckerei under the Federal Foreign Office. The brief: one shared data and AI platform, fully accessible under BITV 2.0, no vendor lock-in, and strict enough to handle everything from crisis monitoring to funding allocation. A tall order for any team. An even taller one inside German procurement.
Building the backbone
As Technical Product Owner Lead, the job was shaping what this platform actually looked like. That meant a containerised multi-tenant architecture with network and identity isolation per ministry. A data layer combining lake, warehouse, contracts, lineage, and catalog. Self-service workbenches: Apache Superset for no-code dashboards, GitLab for inner-source collaboration, ML/Jupyter toolchains for data science teams. Virtualised GPUs with quota policies kept AI workloads elastic without runaway costs. Compliance wasn't bolted on. BITV 2.0, GDPR/DPIA patterns, auditable access logs, and BSI-aligned controls were baked into every layer from day one.
Making it usable, not just compliant
A platform nobody uses is just expensive infrastructure. Analysts, data scientists, and policy advisors each got tailored workflows: governed data onboarding with PII handling, no-code dashboards mapped to ministerial KPIs, and reusable container templates that cut use-case rollout from weeks to days.
Sharing without losing control
Cross-ministry reuse was the whole point, but sovereignty meant every data exchange needed clear boundaries. GitLab group standards and code ownership policies created the guardrails. Inner-source patterns let teams share dashboards, ETL jobs, and ML components. Data-sharing agreements and scoped tokens kept access controlled and auditable.
Safe space for experimentation
Sandbox tenants with synthetic datasets and quota guards gave teams room to explore without risk to production. A pattern library for MLOps, evaluation checklists, and rollback plans made experimentation repeatable, not reckless.
Trust as a system, not a promise
Security governance went beyond access controls. Role-based permissions with just-in-time approvals, periodic access reviews, open-source intake policies, CVE response playbooks, and SBOM audits. Change governance ran through the PLAIN Program Office and security leads jointly.
What it delivered
Version 1.0 went to production in June 2023, roughly eight months after development started in autumn 2022, with Version 2.0 planned from late 2024. The platform earned 2nd place in the eGovernment Competition for “Digital Transformation through AI and Modern Infrastructure.” Five ministries now share data products, models, and workflows on a single sovereign, BITV 2.0 compliant platform.
What runs on it today
Pandemic and political crisis monitoring with scenario insights. Climate-aware land management planning. Supply-chain criticality forecasting. Optimised funding programs across domestic and international portfolios. BMZ data products and AI-assisted dashboards. Each one built on shared infrastructure that didn't exist two years prior.
