Spreading the load: How Salesforce met Multi-AZ HA with SageMaker Inference Components

Salesforce enhanced the availability of its Agentforce AI foundation by using Amazon SageMaker Inference Components with the new SchedulingConfig parameter to ensure multi-AZ high availability and meet compliance requirements. This approach also reduced infrastructure costs by 8x through shared GPU usage while balancing model deployment across availability zones to avoid single points of failure.