A data scientist used SageMaker to train and deploy a loan-default prediction model. The training was done by manually extracting data and running steps in a SageMaker Studio notebook. Prediction accuracy is drifting downward over time. Which combination of steps is the most operationally efficient way to maintain model accuracy? (Choose two.)

The combination of SageMaker Pipelines and SageMaker Model Monitor provides a robust and automated solution for maintaining model accuracy. SageMaker Pipelines automates the entire MLOps workflow, from data preparation and model training to deployment, ensuring that the model can be retrained and redeployed efficiently with fresh data. SageMaker Model Monitor continuously analyzes model predictions and actuals to detect data drift and concept drift, which are common causes of declining accuracy. By setting an accuracy threshold and linking a CloudWatch alarm to the SageMaker Pipelines workflow, retraining can be automatically triggered when accuracy drops, minimizing manual intervention and ensuring the model remains up-to-date. Storing predictions in S3 and running a daily processing job is a reactive approach that lacks automation for retraining. Manually rerunning notebook cells is inefficient and prone to human error. Exporting notebook code to an ECS task with a Lambda trigger is overly complex for this specific problem and doesn't inherently address continuous monitoring or automated retraining.