As an ML Engineer, you need to improve the performance of a batch-processing ML job on AWS that is currently constrained by processing power and memory. Which AWS service should you implement to automatically scale the compute capacity based on the workload demands?