Private AI deployed at your premises
A project to supply and configure servers and run AI models within your organization’s network, against requirements defined with your team. Start with usage and site assessment, then receive a scoped hardware, software, and support proposal.
What determines the hardware?
Model size, document length, concurrent users, and target response time determine capacity. Test a representative workload before choosing compute and memory, then review power, cooling, networking, and installation space. Hardware is not selected on model name alone.
Installation and access
Scope can include model serving, chat, source indexing, authentication, permissions, audit logging, and training. Each item is specified alongside updates, backups, restoration, and server administration responsibilities.
Can it be disconnected from the internet?
An air-gapped deployment can be assessed where needed, but that requires a compatible design for updates, licenses, and external connections. Equipment being on your premises does not by itself establish an air gap.
Commercial structure
A proposal covers hardware, installation, and software, with scoped support and maintenance terms. Hardware, data, code ownership, and model licensing are explicitly addressed in the agreement.