Local AI servers: language models on your own hardware

You use AI on your own hardware, on your premises or in our partner's data centre in Germany, without inputs and documents going to an AI provider. You get the hardware delivered, set up and supported in ongoing operation by conpro IT.

Icons for processor, database and laptop in hexagons on a blue background

AI that keeps your data in-house

Contracts, patient data, design documents or client files often should not reach a cloud provider in the first place. A local AI server processes your inputs and documents in your own network. Your employees still work with a familiar chat interface, ask questions about internal documents and have texts summarised or drafted.

We also tell you openly when a local system is not sufficient or not worthwhile. Whether local, cloud or a combination suits your tasks is clarified by the decision guide Cloud AI or local AI.

What you get from conpro IT

  • Advice on the right size Based on the number of users, tasks and models, you get a recommendation, from the compact desktop device NVIDIA DGX Spark and the DGX Station to a GPU server from Supermicro or Dell.
  • Delivery and set-up conpro IT delivers the hardware and sets up the operating system, language model and chat interface, tailored to your environment.
  • Connection to your documents Manuals, policies and documentation are connected so that the AI can answer questions about them. Connections to specialist software are planned separately.
  • Security and user management The server is integrated into your network and secured. You define who uses the AI and which documents it accesses for whom.
  • Ongoing operation conpro IT operates the system for you, monitors it, installs updates for the system and models and plans the backup with you.
  • Honest assessment Before you buy, you find out whether a local model achieves the quality your tasks require, and where a cloud service is the better choice.

From desktop device to server

Individual specialists and small teams work with a compact system such as NVIDIA DGX Spark that fits on a desk, for example for analysing confidential documents in one department.

Departments with higher requirements use a more powerful DGX Station when larger models or more simultaneous requests are needed.

Several teams or the whole company need a GPU server from Supermicro or Dell, in your server room or in our partner's data centre in Germany. We determine which class is right based on the number of users, the models and your data.

Related topics

Frequently asked questions about local AI servers

Does our data really not leave the company?

Inputs and documents are processed on the server in your network. The system only needs an external connection for updates and remote support, and we set these up in a controlled and traceable way.

Which language models run locally?

Models whose vendor permits operation on your own hardware run locally. Which model is suitable depends on the task, the language and the performance of the hardware. We test it in advance with examples from your day-to-day work.

Who takes care of operation?

conpro IT. You get the hardware delivered and set up by us, and we run the system in ongoing operation with monitoring, updates and dedicated contact persons.

Is a local AI server worthwhile for us?

That depends on how many people use the AI, how confidential your data is and what quality your tasks require. If a local system is not worthwhile, we tell you so and show you a suitable alternative.