# Private AI deployment — Genyra

> Private AI deployment: self-hosted LLMs, retrieval and audit logs on-premise or in a dedicated environment. Residency by design, no training on your data.

Canonical page: https://genyra.com/ai/private-deployment

Models run inside the client’s environment or a dedicated one, with residency, access and logging decided before the first prompt. Custom AI engagements are contracted with Genyra L.L.C-FZ, the separate Dubai company that operates genyra.ai. Genyra.com describes the practice; the agreement names the company doing the work.

## Models run inside the boundary.

The model runtime, the retrieval index, the document store and the audit log run inside the client’s own environment or in a dedicated environment built for the engagement. Questions and documents are not sent to a shared service outside it. The environment can be the organisation’s infrastructure, a private environment we design, or a combination agreed in writing.

The engagement is contracted with Genyra L.L.C-FZ, the Dubai company that operates genyra.ai. Genyra.com describes the practice; the agreement names the entity, the environment and who administers it.

## Residency stated as a design decision.

Where documents may be stored, where processing may happen and which jurisdictions are excluded are recorded as requirements, then met by the design: placement of compute and storage, routing, and the list of permitted outside services, which may be empty. The design is reviewed against those requirements before the system is used.

## No training on client data.

Client documents, questions and answers are used to retrieve and to answer, and for the organisation’s own evaluation. They are not used to train or fine-tune a model unless the organisation asks for that in writing, and then only inside its boundary. The agreement says so; the design makes it so.

## Access and logging.

Ordinary users, service identities and administrators are separated. Users reach the system through the organisation’s identity service; administrators reach the environment through an audited path, never directly. Every question, the passages retrieved, the answer, the reviewer’s decision and the model version are logged inside the boundary, in a store the organisation controls.

## The boundary is drawn before the first prompt.

The list of components inside the environment, the crossings permitted and the data that may pass each one is written down and agreed before any document is indexed or any question asked. Changes to that list go through the same change control as the system itself.

## Bring the workload and its restrictions.

Tell us the workflow, the documents, the environment it must run in and the rules on residency and access. We will discuss a design, the infrastructure it needs and the operating responsibilities. Available hardware and hosting arrangements are confirmed for the particular environment.

## Questions we are asked

**Where does the model run?** Inside your own environment or in a dedicated environment built for the engagement, with the retrieval index, document store and audit log inside the same boundary.

**Can certain jurisdictions be excluded?** Yes. Residency is recorded as a requirement and met by design: placement of compute and storage, routing, and the list of permitted outside services.

**What hardware is needed?** It depends on the models and the workload, and is confirmed for the particular environment. Available hardware and hosting are discussed in an enquiry rather than listed as a standard product.

**What is private AI?** An arrangement where the model runtime, the retrieval index, the document store and the audit log all run inside an environment the organisation controls, with the permitted crossings written down before the first document is indexed.

**Can we self-host an LLM?** Yes. Open-weight models run entirely inside your environment. Which models are practical depends on the accelerator memory available and the workload, which we size for the particular environment rather than from a table.

## Related work

- [Custom AI solutions](https://genyra.com/ai/custom-solutions): Custom AI solutions built around a task your people already perform: contract review, claim triage, case summaries and answers from your records, sources cited.
- [Managed AI operation](https://genyra.com/ai/managed-operation): Managed AI services for deployed systems: an operator, monitoring, version history for models and prompts, evaluation, change control and incident planning.
- [Private environments](https://genyra.com/infrastructure/private-environments): Private cloud architecture and security for applications and data that need deliberate placement: public vs private cloud, isolation, access and administration.

Contact: contact@genyra.com

Genyra LLC and the Dubai business Genyra L.L.C-FZ / Genyra.ai are distinct entities. Project scope and the contracting entity are agreed explicitly.