Amazon SageMaker Unified Studio now supports data profiling and anomaly detection
AWS adds data profiling and anomaly detection to Amazon SageMaker Unified Studio, enabling automated statistical analysis and drift detection for datasets at rest and in transit.
Amazon SageMaker Unified Studio now includes data profiling and anomaly detection features powered by AWS Glue Data Quality. Users can generate statistical profiles of datasets to assess completeness and distribution, while anomaly detection identifies deviations from historical patterns without requiring predefined thresholds. The capabilities apply to both stored catalog tables and data processed in Visual ETL jobs, offering flexibility across workflows.
A new Data profile tab in the catalog interface provides on-demand and scheduled profiling, generating dataset-level and column-level statistics. As historical profiles accumulate, the system builds a baseline of expected behavior, automatically flagging anomalies when data points fall outside predicted ranges. This approach is particularly useful for dynamic datasets where fixed rules may become outdated over time.
For data in transit, the same profiling and anomaly detection capabilities are available on the results page of any Visual ETL job that includes an Evaluate Data Quality transform. This ensures consistent quality monitoring throughout the data pipeline, from ingestion to processing.
The features are available in all AWS Regions where Amazon SageMaker Unified Studio is offered. For additional details, users can refer to the official documentation on the Amazon SageMaker Unified Studio website.