An ML engineer needs to extract, transform, and load data from Amazon S3 for analytics, discover the data schema, and store metadata with minimal manual effort. Which option accomplishes this?

AWS Glue is a serverless data integration service that is purpose-built for ETL operations. It can directly crawl data in Amazon S3, automatically discover schema, and populate the AWS Glue Data Catalog with metadata. This minimizes manual effort and provides a centralized metadata repository for analytics. Amazon SageMaker Data Wrangler is primarily for data preparation and feature engineering, not for robust, large-scale ETL and metadata management. Storing metadata in an S3 bucket lacks the structured, queryable nature of a data catalog. Amazon Athena with AWS Step Functions can run ETL, but Athena is primarily a query service, and it doesn't inherently provide schema discovery and a managed data catalog like Glue. Storing metadata in S3 is not ideal. An Amazon EC2 instance with scikit-learn is a general-purpose compute solution and requires significant manual effort for ETL, schema discovery, and metadata management, which goes against the requirement for minimal manual effort. Amazon Redshift is a data warehouse, not a metadata catalog for S3 data.