ecs-genai

Use whenever someone runs a GPU / ML / GenAI / LLM workload on Amazon ECS: GPU on ECS, ECS GPU-optimized AMI, g4dn/g5/g6/p4/p5 on ECS, which ECS launch type for GPU, Inferentia/Trainium/Neuron on ECS, distributed training, model inference or vLLM/Triton/TGI/Ray on ECS, Capacity Blocks, GPU sharing, ASG per GPU type. Covers GPU on ECS-on-EC2 and ECS Managed Instances (GPU AMIs, NVIDIA runtime, instance families); mixed-instance ASGs supported but constrained (no weighting; managed scaling protects on the smallest type); Capacity Blocks; inference/serving; Neuron; distributed ML; GPU observability; a GPU/ML security slice (PCI/HIPAA). AWS Fargate has NO GPU: use ECS-on-EC2, Managed Instances, ECS Anywhere, or keep Fargate and call Bedrock. Trigger even if GenAI is unsaid. Use eks-genai for Kubernetes/EKS; SageMaker for fully-managed ML; Bedrock for managed foundation models; ecs-architect for non-accelerator ECS design; ecs-build to render the settled design as Terraform; ecs-security for deep compliance.