BSE IT CONSULTING

Sovereign AI · Private RAG · EU Infrastructure

Private AI infrastructure for companies that need to stay in control

Design and build of RAG and LLM platforms on European or customer-owned infrastructure, with controllable data flows, identities, operations and dependencies.

Discuss your AI initiative

Why not just a public AI solution?

For many companies, generic AI services are not a viable path. Not because public cloud is inherently insecure, but because concrete requirements stand in the way:

  • Sensitive internal data and internal access and permission concepts
  • Regulatory requirements and EU hosting requirements
  • Lack of transparency about data flows of external services
  • Vendor lock-in for models, APIs and infrastructure
  • Missing operations and governance structures for AI workloads

What is offered

01

AI Readiness & Architecture

  • Use-case assessment
  • Data and system analysis
  • Target architecture
  • Hosting and model strategy
  • Sovereignty requirements
  • Technical roadmap

02

Private RAG Deployment

  • Document ingestion
  • Embeddings and vector store
  • Retrieval
  • API layer
  • Authentication
  • Access control

03

Sovereign Platform Setup

  • Kubernetes (default: SUSE RKE2) or OpenShift
  • EU cloud or customer-owned infrastructure
  • GitOps
  • Secrets and identity
  • Observability

After the project

Managed AI Operations

After the project, the platform transitions into operations: a monthly operations retainer, included by default in every project proposal. Defined response times within business hours (Mon to Fri, CET) with automated round-the-clock alerting.

Deliberately without 24/7 on-call and without SLA penalties: traceable operations processes instead of formal promises.

Four sovereignty levels

The infrastructure choice follows your actual data requirements, not the other way around.

Level 1

EU region

Data residency in the EU; the operator may be US-owned (e.g. AWS Frankfurt).

Typical: non-critical data with an EU residency requirement.
Limitation: US legal access not ruled out.

Level 2

EU sovereign cloud

EU staff and EU operations, but a US parent company (e.g. AWS European Sovereign Cloud).

Typical: regulated workloads with hyperscaler alignment.
Limitation: the corporate structure remains US-shaped.

Level 3 · Default

EU-owned cloud

No US legal nexus at any layer: Scaleway, STACKIT, IONOS. Default for managed deployments.

Typical: sensitive company data with cloud operations.
Limitation: smaller managed-service portfolio than hyperscalers.

Level 4

On-premises

Model and data entirely on customer-owned hardware.

Typical: highest confidentiality, near-air-gapped environments.
Limitation: hardware, GPU capacity and operations remain with the customer.

The defined entry point

AI Readiness Sprint

A clearly scoped entry point before any infrastructure or implementation decision, with written result documentation as the basis for your decision.

  • One to two workshops
  • Use-case and data analysis
  • Infrastructure and hosting options
  • Privacy and risk questions from a technical perspective
  • Target architecture and roadmap
  • Written result documentation

Fixed price

€2,000

plus VAT

If a project is commissioned within 60 days, €1,000 is credited.

Request the AI Readiness Sprint

Reference architecture and reference platform

A reduced view of a private RAG platform: every component runs on controllable infrastructure.

Reference platform

Own reference platform, verified on four infrastructure targets

BSE IT CONSULTING operates its own reference and development platform for RAG and AI workloads, built to run on different infrastructure targets. The same versioned Helm charts are deployed and verified on:

  • SUSE RKE2 (bare metal)
  • Managed Kubernetes in the EU cloud (Scaleway, STACKIT)
  • AWS European Sovereign Cloud (EKS, Brandenburg region)
  • OpenShift SNO
Argo CD · two-repository pattern Qdrant · document ingestion vLLM · EU-hosted inference Keycloak SSO/OIDC External Secrets · RBAC kube-prometheus-stack · Grafana · Langfuse Terraform

Reference and development platform of BSE IT CONSULTING, and the technical basis of the Sovereign AI offering.

Delivery approach

Phase 1

Use case and requirements

  • Assessed use case
  • Data classification
  • Sovereignty requirements

Phase 2

Architecture and scope

  • Target architecture
  • Hosting decision
  • Clearly scoped project

Phase 3

Pilot or reference implementation

  • Working RAG platform
  • Connected document sources
  • Evaluation with real users

Phase 4

Production readiness and operations model

  • Hardened operations
  • Observability and alerting
  • Defined operations model

Get a technical assessment of your sovereign AI initiative

Briefly describe your initiative, data situation and timeline. You will receive a direct response without any sales layer in between.

Request an intro call