Company

Managed Private Inference, Built in Europe

A Croatia-based technical team configuring, maintaining, monitoring, updating and optimizing private inference inside customer-controlled infrastructure. We pair open model choice with distributor-backed Supermicro procurement and a documented application API.

01 — What we operate

Mission & vision.

The two sentences that shape every decision.

Mission

To make sovereign AI practical through infrastructure customers control and a private inference service they can operate with confidence.

Vision

A Europe whose competitive edge in AI is built on customer-controlled infrastructure, open model choice and local operational expertise. We make that operating model viable.

02 — Where we are

Headquartered in Croatia.

An EU member state, on EU soil, under EU law. Every customer still determines the physical location of its own appliance and data.

Headquarters

Croatia · European Union Member state since July 2013 · Eurozone since January 2023

European startup

Deployment boundary

Managed inside your infrastructure.

The inference environment runs inside the environment you control. Customer applications connect through the documented API and remain outside the supported Product boundary. Nothing leaves your infrastructure.

03 — How we got here

Bootstrapped with product focus.

Building without external capital keeps the team focused on a compact service, clear operating boundaries and capabilities customers can run reliably.

Founders

Three technical co-founders.

Systems engineering, software development and infrastructure operations in one team. We size the environment, coordinate distributor-backed Supermicro procurement and support the service end to end.

Discipline

A compact stack, deliberately chosen.

That discipline shapes how we operate: local inference, a focused Console, an observability layer, lifecycle controls and a documented API contract. Customer application integration remains separately scoped.

04 — Operating principle

Nothing leaves your infrastructure.

The most important sentence in our operating model. It defines the boundary for every customer-premises deployment.

Local observability

Operational metadata only.

The appliance records infrastructure health, model availability, capacity, alerts, update state and metadata-only activity and audit events inside the customer environment. Workload content is not retained by LLM Machines-managed components.

Default data boundary

The data plane is isolated from LLM Machines by default.

User prompts, inference inputs and outputs, application traffic and advanced-integration payloads remain inside the customer infrastructure.

Remote support posture

Remote access is off by default.

Support access is enabled only with explicit customer approval, for a defined time window and with an audit record. Local metrics and alerts remain available to your Operators without opening a standing remote path.

05 — Sovereignty

Built in the EU.
Built for the EU.

Sovereignty isn't a feature — it's the foundation. Every default, every component, every deployment is designed to support regulated European operation, with deployment-specific evidence and review still required.

EU AI ACT — Auditability by deployment

Audit trails, transparency, human oversight.

High-risk AI systems must be inspectable, documented and supervised. Metadata-only activity records, signed audit exports and customer-controlled identity provide a strong operational foundation for use-case-specific documentation and oversight.

GDPR — Data minimisation by design

Personal data stays local by default.

Trade secrets, personal data and confidential records remain inside your infrastructure. LLM Machines-managed components do not retain prompts, responses or request and response bodies.

NIS2 — Cloud-AI supplier risk, reduced

Remove the cloud-AI link from your security perimeter.

NIS2 expanded cybersecurity obligations to essential and important entities and put supply-chain risk on the board agenda. Customer-premises inference reduces dependency on external AI APIs while preserving local identity, observability and operational control.

EU DATA ACT — Portability, by default

Your stack. Your rights. Designed for portability.

The Data Act targets cloud lock-in and reinforces portability and switching rights. Upstream-licensed third-party components, portable models and a documented API contract keep the appliance inspectable and the application boundary familiar.

European team, customer-controlled deployment. Designed in Croatia for European enterprises and operated inside your infrastructure.

SRC

Source-available first-party Product layer. Original first-party LLM Machines Product source is source-available under the unmodified PolyForm Internal Use License 1.0.0. Third-party components retain their upstream licences.

⌂

Deployed inside your perimeter. Your infrastructure, your hardware, your operating controls. Nothing leaves your infrastructure.

06 — Beyond residency

Sovereignty is more than where data sits.

For AI workloads, sovereignty depends on who controls the data plane, the operating model and the technology stack.

Data sovereignty

Your inference stays local.

The appliance runs inside your infrastructure. Prompts, responses and workload content remain within the environment you control.

Operational sovereignty

You decide who can access the system.

Identity, Admin and Operator roles, support access and infrastructure controls remain under your authority. Remote support is off by default and explicitly enabled when needed.

Technology sovereignty

Upstream-licensed components support switching.

Chat Completions is the documented API protocol your favourite apps and harnesses use to access approved local models. Upstream-licensed third-party components and model portability keep your options open.

What's next

Bring managed private inference in-house.

We operate private inference inside your infrastructure. Your applications and harnesses connect through the documented API.

See the technology See pricing Talk to sales →