The Inference Engine
The engine that runs a large model on your own machine. Offline, audited, on hardware you already own.
Sovereign AI has had one stubborn problem. To run a genuinely capable model, you needed a datacentre, and that priced out most of the organisations that need control of their own intelligence the most: banks, hospitals, government, anyone who can't send their most sensitive data to someone else's cloud and trust a policy page. Poros is the fix. It's the piece that turns sovereign AI from an infrastructure project into something you run on your own terms.
It runs local
Poros runs a large model on a single workstation-class GPU. It reads a machine's GPU, memory and CUDA capability and decides which models it can run; only NVIDIA Turing has been validated so far. Poros lets a single workstation GPU work with system memory to run models larger than the GPU's own memory.
It runs sealed
Poros connects only to a loopback address, with no name lookup. It's a loopback-only component, and an independent adversarial review confirmed the egress guard holds. That's why air-gapped is a property you can test here, not a promise on a slide.
It runs accountable
Every session seals one signed record in the customer's own audit chain for every request routed through Poros. Not trust us. Verify it.
Where it stands
Poros the engine is built and independently reviewed, and tested ahead of integration; the installed product does not run on it yet. It runs a large model on a workstation GPU today, with one signed record in the customer's own audit chain for every request routed through it. The full desktop assistant and the agentic studios are the next phases of the build, not something we're claiming as finished.