About
AI infrastructure you can keep inside your boundary.
Federis builds the control plane that orchestrates and secures AI workloads in air-gapped, on-premises, hybrid and sovereign cloud environments, and the studio where teams build agents on top of it.
A sovereign control plane between your applications and your GPUs, and the studio where teams build on top of it.
Federis
A modular sovereign control plane, network controller and security proxy. It sits between applications and GPU-backed LLM serving layers, and it decouples compute and storage from central governance and audit requirements.
Federis Studio
The agent and workflow builder. Run it on its own, or as a module of Federis so that every workflow uses your deployed models, policy, guardrails and audit.
How we think about it
Sovereign by design
Complete control over data residency, security and regulatory compliance, whether the deployment is on-prem, hybrid or at the edge.
Composable
External databases, identity platforms and vector databases connect through interfaces, so you can swap them without touching the core.
Verifiable
Decisions are recorded in a SHA-256 hash-chained ledger under WORM retention, so what happened can be shown, not just claimed.
Frequently asked questions
What is Federis?
A modular sovereign control plane, network controller and security proxy. It orchestrates and secures AI workloads and sits between your applications and your GPU-backed LLM serving layers.
Where does it run?
In air-gapped, on-premises, hybrid or sovereign cloud environments: on a single everyday machine, on bare metal, or on private Kubernetes (v1.28 or later) for clusters. Serving runtimes pull model weights from local read-only volumes, so air-gapped targets need no outside calls.
What does a deployment need?
As little as your model use case needs. The control plane is light enough to run on minimal infrastructure, even an everyday system, and it grows only when the workload does. For a highly available production reference, for example, three control plane nodes (e.g. 8 vCPUs, 32 GB RAM, 200 GB SSD each) and two redundant GPU nodes (e.g. 16 vCPUs, 64 GB RAM, a 1 TB model cache each).
How do we get started?
Three steps: plan the sovereign boundary with our team, bootstrap the platform with declarative GitOps manifests, then choose an operations and support package that fits your team. See the services page for detail.
How does Federis Studio relate to Federis?
Studio builds and runs agents and workflows. In its platform edition it uses Federis for identity, deployed models, policy, guardrails, retrieval and audit. It can also run standalone. See the Studio page.
Talk to the people who build it
Tell us about your environment and what you need to govern.
- Sovereign by design
- Composable
- Verifiable