Developed a flexible historical feature retrieval capability for the red-hat-data-services/feast repository, focusing on the ClickHouse offline store. The work enabled get_historical_features to operate without requiring an entity DataFrame by introducing an end_date parameter, which generates a synthetic single-row DataFrame when none is provided. This approach streamlined data preparation and allowed for precise control of point-in-time join windows and data retention through TTL settings. The solution was validated with integration tests against a real ClickHouse container, ensuring production readiness. The project leveraged Python programming, data engineering, and database management skills, with a strong emphasis on robust unit and integration testing.