How AI and Data rely on Foundation
How the higher platform layers consume Foundation capabilities.
Foundation is the operational substrate for the rest of the BullSequana AI platform. AI and Data build on top of it rather than operating independently from it.
AI on top of Foundation
AI relies on Foundation for:
- API exposure, through the shared gateway and its routes
- authentication and access, through the shared identity and security model
- inference execution, through KServe, the runtime profiles, KAI Scheduler, and the internal model gateway
- workflow execution, through Temporal and the Argo workflow engines
- state and artifact storage, through PostgreSQL, Milvus, and S3-compatible object storage from Rook-Ceph or an external provider
- tenant runtime, through the BSQAI Tenant Operator, which prepares each tenant's namespace, storage, databases, and GPU queues
- observability, through the shared monitoring, logging, and tracing stack
In practical terms, when an AI service is deployed, Foundation provides most of the platform mechanics that make the service operational. How AI builds on Foundation maps each capability to the AI services that use it.
Data on top of Foundation
Data also depends on Foundation for the same concerns:
- secure exposure of user and service endpoints
- identity and access integration
- platform-level observability
- persistent storage services where needed
- GitOps-based deployment and lifecycle management
Data adds enterprise data capabilities, but it still uses the same Foundation layer as the rest of the stack.
Why this layering helps
This layered dependency model makes the platform easier to evolve:
- Foundation standardizes shared concerns
- AI and Data focus on product capabilities
- use cases can be built without re-solving infrastructure questions each time