Scale AI training and inference for drug discovery through Amazon EKS and Karpenter

This article discusses how the startup Iambic Therapeutics used Amazon EKS, Karpenter, and KEDA to efficiently scale AI training and inference workloads for drug discovery on AWS.

Specifically, the article covers:

• Iambic's need for scalable AI training and inference to rapidly design and evaluate millions of drug molecule candidates
• The solution architecture combining Amazon EKS, Karpenter for Kubernetes node auto-scaling, and KEDA for Kubernetes pod auto-scaling based on custom metrics
• How the solution enables efficient utilization of GPU instances and consolidation of workloads to minimize costs
• Step-by-step instructions for deploying the solution
• Visualizations showing the solution's scaling behavior with GPU node provisioning and pod scaling
• Significant performance gains and cost savings achieved by Iambic using this approach for AI training and inference