A machine learning team runs a custom training algorithm on SageMaker that needs external assets. They must submit both their algorithm code and algorithm-specific parameters to SageMaker. Which two AWS services should they use to build and supply a custom algorithm on SageMaker? (Choose two.)

Amazon ECR is essential for storing and managing Docker images. SageMaker training jobs run within Docker containers, so the custom algorithm code, along with its dependencies, must be packaged into a Docker image and pushed to ECR. SageMaker then pulls this image to execute the training. Amazon S3 is used to store the algorithm-specific parameters and any other external assets (like datasets or pre-trained models) that the custom algorithm needs during training. SageMaker can easily access these resources from S3 buckets. AWS Secrets Manager is for sensitive information, not general algorithm code or parameters. AWS CodeStar is a development tool for CI/CD, not for storing container images or assets. Amazon ECS is a container orchestration service, but SageMaker manages the container execution for training jobs, making ECR the relevant service for image storage.