Running AI In-House (Securely & in Compliance with the GDPR)

AI in Business

Operate your own LLMs in compliance with data protection regulations—using vLLM and OpenWebUI, in your own data center or in the NWS Cloud. Your employees use a ChatGPT-like tool without a single document ever leaving the office. You choose the model and location yourself.

Data remains in-house

Neither prompts nor responses leave your infrastructure—ideal for sensitive business and customer data.

GDPR-compliant & based in the EU

Operated in our own data center or in the NWS Cloud in Germany—rather than by a provider outside the EU.

Free choice of model

Open-weight models for control and cost, or frontier models like Claude and ChatGPT for maximum performance—the choice is yours.

Keeping Costs Under Control

Open-weight models provide cost-effective coverage for about 80% of cases—without any unexpected token charges per request.

Complete flexibility

Operate it yourself, use NWS, or combine both—and switch the operating location or model at any time.

Get More Value from Your Own Data

Linked to internal knowledge (RAG), the AI provides concrete, reliable answers instead of generic platitudes.

The Problem

AI should be used—but sensitive data must not be stored in third-party clouds. It is precisely this conflict that slows down many projects, often without people realizing it.

Data Leaves the Company

With ChatGPT, Claude, and similar services, prompts and documents end up with providers outside the EU. Data protection often becomes an issue too late in the process.

Dependency & Costs

Per-token billing and vendor lock-in make costs unpredictable—and you have little freedom to choose your model or location.

Little Added Value Without Context

A generic AI does not have access to your internal data. Without connecting them to one’s own knowledge, the answers remain superficial.

How we work with you

Four steps, the same for every NETWAYS solution—from selecting the right models and hardware to ensuring the stable operation of your AI platform.

Step 1

Analysis & Concept

We'll assess your use cases, data protection requirements, and existing GPU hardware—and select the appropriate models.

→ You'll get the right model size instead of an overpriced, oversized one.

"

Step 2

Setup & Integration

We deploy vLLM as an inference backend and provide a familiar chat interface using OpenWebUI—either in our own data center or at NWS.

→ An OpenAI-compatible API—existing tools can be integrated directly.

"

”Step

”Commissioning

”

"

Step 4

Support & Operations

If you'd like, we can handle all aspects of operation, updates, scaling, and GPU monitoring (MyEngineer)—or we can train your team.

→ Your AI remains stable and up-to-date without the need for your own team of specialists.

Building Blocks of Your Solution

For each module, you decide how much you manage yourself and where you rely on NETWAYS services.

Choose Your Own LLM

Select a model

Open-weight models such as Llama, Mistral, or Qwen for control and cost—or a Frontier model via API for maximum performance.

Result: the right balance between data protection, performance, and cost.

On-premises or EU cloud

Select a location

In your own data center for maximum control, or in the NWS Cloud based in Germany—both are GDPR-compliant and located within the EU.

Result: Full control over data, without having to require customers to use their own hosting.

OpenWebUI & API

Surface & Access

OpenWebUI serves as a familiar chat interface, plus an OpenAI-compatible API that your own applications can connect to directly.

Result: a ChatGPT-like experience for employees—all handled in-house.

Your Own Data via RAG

Connect Your Own Data

Through RAG, the AI accesses knowledge databases, documents, and internal systems—and you retain control over these sources.

Result: Answers based on your actual company knowledge.

What You’ll Achieve

Full control over your data, predictable costs, true independence.

Data Sovereignty

Your data remains within our company or within the EU. Use AI without compromising data privacy.

Predictable Costs

Open-weight models on your own or rented hardware instead of per-token billing—making costs predictable.

Independence

No lock-in: You are free to choose the model, location, and provider, and can switch at any time.

What is your solution built with?

Tried-and-true open-source components. You decide which components you’ll manage yourself and where you’ll rely on NETWAYS services.

vLLM

High-performance inference backend for production-grade LLM operations: high throughput, efficient GPU utilization, and an OpenAI-compatible API—all running on your infrastructure.

OpenWebUI

Self-hosted chat interface for LLMs—a familiar, ChatGPT-like tool with RAG and tool integration that runs entirely offline.

n8n

Connects AI to internal systems via RAG and automation, and triggers real-world actions directly from the chat—all in-house and in compliance with data protection regulations.

Grafana

Keep an eye on your AI platform’s GPU utilization, throughput, and availability—so operations remain predictable and stable.

We’ll integrate what you’re already using with

Your AI platform can be customized and integrated with existing systems. A selection of the components and interfaces we typically work with.

Models & Weights

Llama
Mistral
Qwen
DeepSeek
Gemma

Surfaces & Access

OpenWebUI
OpenAI-compatible API
VS Code / Claude Code
Custom Applications

Operations & Hardware

NWS Cloud (EU)
In-house data center
GPU Server
Kubernetes / Docker
nws.netways.de

Inference & Serving

vLLM
Ollama
NWS AI
Hugging Face
OpenAI / Anthropic (API)

Own Data (RAG)

PostgreSQL / pgvector
Qdrant
Elastic / OpenSearch
Nextcloud
SharePoint

Questions & Answers

Frequently Asked Questions About This Solution

Which AI is GDPR-compliant?

2

3

AI is considered GDPR-compliant above all when the data remains under your control. The surest way to ensure this is to run the model in your own data center or in an EU cloud such as NWS, so that no prompts or responses are sent to providers outside the EU. NETWAYS sets up exactly this kind of seamless operation.

Is ChatGPT GDPR-compliant?

2

3

When using ChatGPT, Claude, and similar services directly, user input typically leaves the EU—which, depending on the type of data, can be problematic and often does not automatically comply with data protection regulations. Those who wish to use Frontier models can do so selectively and only for non-sensitive data; sensitive content is processed by a self-hosted OpenWeight model on-premises.

What is on-premises AI?

2

3

On-premises AI means that the language model runs on your own hardware in your own data center—rather than as a service in a third-party cloud. You retain full control over the data, the model, and operations. To this end, NETWAYS uses vLLM as the backend and OpenWebUI as the user interface.

How do I use AI in compliance with data protection regulations?

2

3

By carefully choosing your location and model: Open-Weight models hosted in your own data center or in the NWS Cloud in Germany keep your data on-premises. A RAG integration provides your own content, while keeping the sources under your control. Here's how to use AI productively without revealing data.

Do I need expensive models, or are open-weight models sufficient?

2

3

For most use cases, open-weight models such as Llama, Mistral, or Qwen are perfectly adequate—they cover about 80% of cases at a low cost. Frontier models are particularly worthwhile in situations where maximum performance is required. You can combine both approaches and decide which one to use depending on the task.

What are vLLM and OpenWebUI?

2

3

vLLM is a high-performance inference backend that efficiently runs open-source LLMs on your GPU hardware and provides an OpenAI-compatible API. OpenWebUI is the self-hosted chat interface for this—a familiar, ChatGPT-like tool with RAG and tool integration that runs entirely offline.

Book a personal consultation with ChristianIndividual open source solutions tailored to you and your business.Get in touch