Skip to content

Deployment and data

Choose where your AI runs and how data flows.

Compare SaaS, using the solution in a managed environment, with assisted deployment in the company’s infrastructure. The choice defines who administers each part and needs to account for every data path.

How to buy

Product and purchasing channel.

Shared SaaS or Dedicated: two deployment options. Dedicated can use OCI, AWS or on-premises infrastructure, subject to technical assessment and the agreed scope. Talk to Starya about purchasing directly or through AWS Marketplace and Oracle Marketplace. The purchasing channel is defined separately from the environment where your operation runs.

Explore Shared SaaS, Dedicated and purchasing channels ↗

Deployment options

Understand what changes for each team.

Use the solution in a managed environment.

SaaS means software as a service. The company uses the environment managed by Starya, with users, data and integrations defined for the operation.

Environment
Infrastructure managed by Starya, according to the agreed solution.
Deployment and evolution
Starya organizes deployment and updates for the contracted environment.
The company's role
Define responsible people, users, authorized sources and usage criteria.
Data and models
Region, storage, providers and data flows need to be defined in the design.
Compare both options side by side
Responsibilities by deployment option
AspectManaged SaaSCustomer infrastructure
EnvironmentInfrastructure managed by Starya, according to the agreed solution.Cloud account, virtual private cloud (VPC) or your own infrastructure, depending on feasibility and scope.
Deployment and evolutionStarya organizes deployment and updates for the contracted environment.Joint planning of installation, updates, monitoring and support.
The company's roleDefine responsible people, users, authorized sources and usage criteria.Manage the infrastructure and access under its responsibility; designate the technical team.
Data and modelsRegion, storage, providers and data flows need to be defined in the design.Define what stays local, what may leave and which providers may be used.

Reference scenario · not a product configuration

Where the model processes data
Explore data flows

Starya-managed environment

Operation

Agents + team

Context

Authorized sources

Records and copies

Destinations, retention and support to be defined

External destinations

Model provider

Call content ⇄ response

External system or channel

Action data ⇄ confirmation, when provided by the integration.

Support and associated services

Authorized records and access. Flows to be defined per component.

Model call

Only content needed for the call should enter this path. Provider, region and retention must be defined.

A local model and company-owned infrastructure do not automatically guarantee sovereignty. The scenario must account for all outbound flows and access.

Options are subject to technical assessment. Customer deployment is assisted; components and support vary by scope. This does not imply a standalone installer or offline operation.

Data sovereignty

Define where data resides and who can use it.

Data sovereignty involves control over storage, processing, access and movement. These decisions need to cover the services used by the operation.

Where data resides

Which regions and environments hold documents, records, copies and search indexes?

Who can access it

Which teams, services and providers have access? Who grants, reviews and revokes permissions?

Where it is processed

Which models and services receive content? Are local options or authorized external providers available?

What can move

What content or metadata can leave the environment? How is this recorded and verified?

Your own infrastructure requires a complete design.

If an agent calls an external model or service, some content may leave the environment. Restrictions need to account for every component used by the operation.

Federated AI · reference architecture

Operations across environments. Shared coordination.

In this design, federated AI means coordinating operations that run in different units or environments. Each keeps its own owners and sharing boundaries. Explore the reference architecture.

Evolving reference architecture

Shared coordination

Versions, rules and monitoring according to the design.

What a site shares

↑ Execution status · ↓ version and rule

Environment A · operation status

The design shares status and version without requiring the full document contents. Even these records may contain information requiring restrictions.

Conceptual reference, subject to technical feasibility. This diagram does not demonstrate available federation or one site’s access to another’s data.

Reference design for technical discussion. It does not demonstrate general availability or isolation of all data in a deployment.