An application user sends a query. The web servers deliver to the query service. At this point, the query service can employ machine learning (ML) models using Amazon SageMaker (arrow not shown) for user segmentation, concept and entity extraction, query-to-click, and other data to enrich the query. The query service enriches or rewrites the query, based on user segmentation from Amazon SageMaker (arrow not shown), user preferences from Amazon Relational Database Service (Amazon RDS), and past query performance. It sends the augmented query to Amazon OpenSearch Service. The user sends only searchable data to Amazon OpenSearch Service, employing a relational or NoSQL system as the system of record. The query service retrieves only keys in the search results. It retrieves the full record information from the system of record. The web servers and query service send user interaction data back to an Amazon Simple Storage Service (Amazon S3) data lake or Amazon Redshift. An offline process pulls user interaction from the data lake. The offline process takes data (such as clicks) that it needs to augment the
