Training and tuning
Accelerators, distributed training, fine-tuning and evaluation workloads use capacity appropriate to the job.
STATIC AI Cloud
Accelerated infrastructure for training, serving and running AI workloads.
INSIDE STATIC AI CLOUD
Purpose, systems, and connections.
AI Cloud supplies the computation and supporting infrastructure for model development and operation. It keeps the model portfolio, model selection and the user’s relationship with SAGE separate from the underlying capacity.
Serving and retrieval
Real-time and batch inference connect to model serving, model artifacts and vector or retrieval infrastructure.
Capacity choices
Reservations, AI networking and approved outside compute can support workloads before or alongside STATIC-operated infrastructure.
How it connects
Running a workload does not grant permission to train on the customer’s data or transfer ownership of their model weights.
Models owns the model portfolio; the model router selects eligible models; Cloud provides execution.