Saptiva AI runs wherever you decide. The same applications, the same orchestration, the same audit surface, across every deployment mode. When your infrastructure changes, the platform follows. You don’t replatform.
A capability of frIdA.
Public cloud. Private cloud. On-premise. Air-gapped.
Deployment mode is a runtime concern, not a product concern. The applications, the orchestration engine, the routing configuration, and the audit surface are identical across public cloud, hybrid, on-premise, and air-gapped deployments. Only the compute placement changes.
Every Saptiva AI deployment sits in one (or more) of these four modes. frIdA routes workloads across them according to the configuration you set. A single customer deployment routinely runs workloads in multiple modes simultaneously, against different data classes.
Elastic compute for non-resident workloads.
Non-sensitive workloads, general-purpose tasks, bursting capacity, development and staging environments. Not suitable for sensitive data flows in most jurisdictions.
Workloads carrying sensitive data are placed elsewhere by configuration.
In-country cloud for sensitive workloads.
Regulated data at volume, multi-tenant banking workloads, KYC at scale, customer operations with data residency requirements.
Residency set by configuration. Egress controlled at your perimeter. Customer-managed encryption keys. Full audit trail retained in-country.
Sensitive data classes run inside your perimeter. Everything else routes to the cheapest capable compute - frontier models, burst capacity, a neocloud - under the same configuration and the same audit trail. One control plane. Two environments. One record.
Enterprises with a perimeter to protect and a budget to protect. MultiMoney runs it today: on-premises and private cloud, across six countries.
DELOITTE, AUG 2026: THE HYBRID PARADIGM - A CENTRAL CONTROL PLANE GOVERNS IDENTITY, AUDIT AND COST WHILE EXECUTION STAYS DISTRIBUTED.
Customer data center, customer hardware.
Institutions with an on-prem mandate, workloads that cannot leave customer-controlled infrastructure, and the highest-sensitivity financial data flows.
Network egress controlled at the customer’s perimeter. Audit stream consumed locally. Model weights resident on customer hardware.
Disconnected environments, maximum isolation.
Critical infrastructure and the most sensitive financial workloads. Environments where the connectivity requirement itself is a risk.
Zero external connectivity at runtime. Model and configuration updates via authorized, signed offline transfer. Full audit surface available locally.
A single deployment typically uses more than one mode, with different data classes routed to different environments by the configuration frIdA applies. Common routings across enterprises.
Most enterprise AI platforms treat deployment mode as a rebuild event. Moving from cloud to on-prem means new code, new configuration, new audit surface, new certification. Saptiva AI was built from the start to treat mode as a runtime concern.
Update the routing configuration. frIdA moves the affected workloads on the next run. The application code does not change. The audit record remains continuous.
Add the country’s environment to the configuration. The same Studio applications deploy against local compute. Two countries, one codebase, two configurations.
frIdA’s compute target changes by configuration. Zero application rewrite. Zero model retraining. Audit record remains portable.
Install the air-gapped runtime at the customer site. Route the specific workload by configuration. The other workloads are unaffected. The platform remains one platform.
Add the cloud target to the configuration. Sensitive classes stay pinned inside; the rest routes out on the next run. One line changes. The record does not.
Every deployment mode relies on infrastructure we do not sell. Our partners provide the hardware, the cloud capacity, and the in-country distribution. Saptiva AI orchestrates above them.
Hybrid rack you’re still standing up. In-country region you’re migrating to. Isolated cluster your security team insists on. Saptiva AI is already there. A Forward Deployed Engineer will respond within 48 hours.