On-premise & hybrid: the operating model follows the nature of the data.
Four operating models, one platform. The data class decides which task runs where – not a dogma for or against the cloud.
In brief: In hybrid operation, sensitive tasks can run on your own servers or in a Swiss private cloud. External models can be used for non-critical tasks. We check suitability by data type, contract and integration.
Operating model per task.
| Task | Data class | Recommended model | Operating model |
|---|---|---|---|
| Contracts, HR data, patient administration | highly sensitive | local | On-premise / private cloud |
| Internal knowledge search | internal | local | On-premise / private cloud |
| Meeting transcription | internal | local or private cloud | Private cloud / hybrid |
| Web research, market monitoring | public | external | Hybrid |
| Bulk classification of non-critical emails | low | external or small local | Hybrid |
What on-premise requires.
- Compute nodes with a GPU or enough CPU capacity for the chosen models
- Network, storage, backup and access control
- Operational responsibility or operational support from Swiss AI Labs
- Updates for models and platform
- Ongoing monitoring of quality, costs and utilisation
Define data classes and model access.
Hybrid only works if the assignment is clear. Data classes are defined in Studio, and Swiss AI Factory is designed to enforce them during routing. A task with highly sensitive data must not go to an external model; the rule is configured and monitored.
Compliance also depends on purpose, contract, data location, technical measures and ongoing oversight.
Related topics.
Frequently asked questions about on-premise & hybrid
Can we start in the cloud and move on-premise later?
A decoupled architecture can make the switch easier. Compute nodes, interfaces, contracts, testing and adjustments still need to be checked.
Who runs the compute nodes?
You, your IT partner or Swiss AI Labs as part of the agreed operational support.
Do you show internal details?
Publicly, we show the principle, layers and data flow at a high level. Internal topology, routing rules, security rules and version states stay internal.
Is it already in production?
We state the maturity of each building block openly. Phrases such as “is designed to” describe the architecture, not a client deployment. Studio, Agents and parts of Factory run in our own development and operating environment and are continuously extended. Further operating models, multiple nodes and private and local installations will follow step by step.
The right operating model.
We map your tasks to data classes and operating models.