A data science team has built a SageMaker Pipeline for model training and evaluation. After the model passes evaluation, they need the pipeline to notify an external inventory management system by calling its REST API and to record the model version in a DynamoDB table. These actions take only a few seconds and require no ML compute. The team wants to keep this logic inside the pipeline definition without provisioning any long-running or ML-specific infrastructure. Which pipeline step type should they use to implement these actions?