Fundaments Labs: Private AI Workbench

Fundaments Labs

In this series, we would like to introduce a Fundaments Lab project: The Private AI workbench offers an easy entry point for any user who wants to apply an AI model. So far, nothing new, as many (public) cloud providers already offer this. But what if this AI needs to be used on sensitive, business-critical data? Then this must be done in a controlled manner, without anyone or anything (another AI model) getting hold of this data. With the Private AI Workbench that Fundaments is developing in this LAB, private &  soevereine Cloud comes together with the application of AI: a 100% guarantee of data protection from a technical, legal, and operational perspective. Your data in a Dutch cloud, fully in line with Dutch and EU legislation and managed completely and transparently by a Dutch organization.

‍

Private AI

The Private AI Workbench is designed to enable you to process your data with AI in a private cloud solution. This concept ensures that your data is used in a secure, controlled environment. The environment is rolled out from Fundaments' data centers and is not connected to the public internet, which means it does not use AI model services as we know them, such as ChatGPT / Gemini / etc. This ensures that your sensitive data is processed under your control and Fundaments' supervision.  

By using our own private cloud infrastructure, optimized for AI processing, your sensitive data can be both trained and combined in Retrieval Augmented Generation (RAG). This is shown schematically in the figure below.

‍

What exactly does this Private AI Workbench offer?  

In the Private AI Workbench, Fundaments offers a concept in which anyone who wants to work with AI models can access a private cloud with AI-optimized hardware. All of this is very accessible, both in terms of ease of use and cost.

It is structured as follows.

• Private Cloud: From one of Fundaments' data centers, we offer physical server hardware that is optimized for the use of AI models. NVMe storage for fast loading and switching between models, and of course a GPU for processing AI model transactions (tokens). This hardware is then virtualized and equipped with the workbench tools: AI models, an interface, and access to the entire environment, allowing you to work in isolation, as in a ‘sandbox’.  

• For access, a VPN with an encrypted connection is provided, so that the private cloud is fully protected and accessible.  

• The Ollama framework, which allows various (language) models to be installed and configured; these are delivered ready to use, so you can get started right away.  

• OpenWebUI / LibreChat: this user interface allows you to deploy AI models, but also offers the possibility to load and modify models. Furthermore, this interface also provides options for storing files and having them processed by the AI model.  

• N8n: to deploy AI models automatically, the Private AI Workbench offers workflow automation. Here, data can be processed by one or more AI models. An ideal starting point for your first automation with AI.

‍

‍

Evolution of the Private AI Workbench  

The Private AI Workbench is designed to make it easy to harness the power of AI and roll out proof of concepts that are accessible and inexpensive. At Fundaments LABS, we are looking for parties who want to contribute their ideas and help us further develop the workbench.  

• Are you developing an application that requires AI support?  

• Are you a start-up looking for a scalable AI solution?

• Or do you want complete control over your data in a data model of your choice?

Get in touch with us so we can share our Fundaments LABS with you.

‍

‍