Private AI, deployed into a national procurement body and a state seaport

The problem

Public bodies, regulated companies and companies that really care about their IP want what ChatGPT gives their staff, but they cannot send procurement documents, port operations data or internal files to a public API. They also need to control what the tools cost, who used them for what, and whether they could walk away from a vendor without losing everything.

We kept hearing the same requirements from different clients. So we built the platform once, on open-source foundations, and deploy it into each client’s own cloud.

What we built

VAIOS is a private, multi-model AI platform. The core is open source – Open WebUI for the chat interface, LiteLLM as the gateway to more than 200 models, Langflow for agents and workflows – and Vilantis adds what organizations actually need around it:

• A SharePoint indexer that turns a document library into retrieval-augmented generation (RAG): staff ask questions, the system answers from the organization’s own documents and shows its sources.
• Cost governance: per-user audit trails, spend dashboards, alerts and hard limits.
• Embeddable assistants for intranets and public portals, and client-specific chatbots and integrations built on the platform.
• Deployment into the client’s own Azure subscription, with the full source code and the right to keep it if the partnership ends.

How organizations run it

VAIOS runs today at CPO, Lithuania’s central purchasing organization; at the Port of Klaipėda, the state seaport authority; and at a manufacturing company that prefers not to be named. Each runs its own instance in its own cloud. Vilantis maintains the platform, builds the integrations each client needs, and hands over the code.

The pattern carries over to any company that cannot send its documents to a public AI app: governed, private AI on your own infrastructure, built by a team that stays to run it. VAIOS has its own site for European buyers at vaios.lt.