Xenith Private AI
Open-weight LLM stack inside your VPC or air-gapped network.
Xenith Private AI is an open-weight inference stack (vLLM, Llama 3 or Qwen class models, BGE embeddings) deployed inside your VPC or air-gapped network for legal, banking, healthcare, and defence data that cannot leave the perimeter. A model gateway with per-team keys, quotas, and prompt logging fronts the stack; agents, RAG, and assistants connect through it without code changes. Architecture is mapped to SOC 2 and HIPAA controls; your auditors own the certification.
Deployed as a scoped engagement for your stack - not a self-serve trial or live sandbox.
How it works
- Full Xenith stack on your servers. Nothing phones home.
- Air-gapped operation supported, works with no internet connection
- Open-source LLMs (Llama, Mistral, Qwen) deployed and optimised for your hardware
- SOC 2 and HIPAA-aligned architecture, compliance documentation included
- Dedicated Xenqube engineer for the first 90 days post-deployment
Who this is for
What's included
Typical engagement deliverables
Scoped to your stack and compliance needs. We deploy the working system into your environment and hand over runbooks - not a sandbox login.
Infrastructure assessed and deployment plan approved by your team
LLM running on your hardware, first queries responding
Security architecture documented and reviewed with your compliance team
Admin panel for managing models, users, and access live
Runbook for your ops team to manage the system post-handover
Questions buyers ask
Frequently asked questions
- When does on-premise beat API pricing?
- At sustained high daily token volumes with steady load. Below that, a private endpoint in your own cloud account usually wins. We deliver a TCO memo with the break-even and the assumptions you can change; the on-premise math post shows the model.
- Are open-weight models good enough?
- For classification, extraction, drafting, and retrieval-grounded answers, Llama 3 and Qwen class models pass most eval sets we have run. A frontier model can stay available through the same gateway under a stricter data policy for hard reasoning tasks.
- Can it run air-gapped?
- Yes. Weights, containers, and dependencies ship as an offline bundle and updates follow your change window. The stack makes no outbound calls.
- Who owns HIPAA or SOC 2 certification?
- You do. We design data flows, access controls, and audit logs to the control requirements and hand over the evidence. Your auditors certify.
When this is the wrong fit
- Teams with no residency or compliance driver; API models with a data-processing agreement are cheaper.
- Organisations with nobody to operate GPUs or Kubernetes after handover, unless a retainer is in place.