Annie is infrastructure-agnostic. It is designed to deploy across a range of options to meet your sovereignty, performance, and commercial requirements:
- On-premise — your own data centre hardware, fully isolated from external networks
- Australian sovereign cloud — certified Australian cloud providers hosting workloads under Australian jurisdiction, with data that never leaves Australian soil
- Private cloud — your organisation's private cloud environment, separated from public cloud regions
- Hybrid — a combination of on-premise and sovereign cloud for different pipeline roles or redundancy requirements
Annie does not require any connection to Evari's infrastructure after initial deployment. Model weights, configuration, and fine-tuning data all reside on your infrastructure. The only external dependency is optional telemetry, which can be disabled for fully air-gapped deployments.
Yes. Annie is designed to operate with zero external network connectivity after initial deployment.
In an air-gapped configuration:
- All model inference runs on your hardware with no outbound API calls
- The Cognition Stream runs on your infrastructure using only data accumulated locally
- There is no dependency on Evari's systems, external AI APIs, or third-party services for any operational function
- Updates (model weights, platform software) are delivered via secure offline transfer on customer-defined schedules
Air-gapped deployment is the default configuration for government and defence customers, and is supported at all deployment tiers.
If you require ASD ISM-aligned deployment guidance or are building toward Australian Government PROTECTED classification, raise this during your enterprise briefing.
Yes. Annie is designed to run on Australian sovereign cloud infrastructure — cloud environments where data is stored and processed within Australian jurisdiction, under Australian law, with no cross-border data transfers.
Annie's architecture supports this natively: model inference, fine-tuning, and all operational data remain within the environment you specify. Annie makes no external API calls that would cause data to traverse to overseas servers.
For financial services customers subject to APRA CPS 234, and for government customers subject to ASD ISM requirements, Australian sovereign cloud deployment supports the data residency controls these frameworks require.
Contact us to discuss which certified Australian cloud platforms are supported for your sector and the specifics of your residency requirements.
A production Annie deployment typically takes four to twelve weeks, depending on infrastructure complexity and the number of pipeline roles being configured at launch.
A typical deployment sequence:
- Infrastructure selection — identify and confirm your deployment environment (on-premise, sovereign cloud, hybrid)
- Foundation model selection — configure which Workforce Foundation Models and/or your own models fill each pipeline role
- Domain configuration — connect Annie to your data sources, configure task types, and run initial validation against your workflows
- Validation and go-live — run acceptance testing with your team, configure Cognition Stream schedules, go live
The enterprise briefing process identifies your requirements and produces a deployment plan before any commitment is made. Organisations with existing GPU infrastructure and a clear use case typically reach the shorter end of the timeline.