Whenever you're ready, walk me through how you'd approach this.
What is the role of data preprocessing in machine learning?
Explain how you would design preprocessing as a repeatable pipeline for training and inference. Cover data cleaning, missing values, categorical encoding, numerical scaling, outlier handling, feature consistency, leakage prevention, validation, and monitoring. Describe how you would implement the workflow with Amazon SageMaker Processing, AWS Glue, Amazon S3, and a model-serving endpoint, including how transformations are versioned and reused across environments.