Hopsworks Query Service provides a Python API that enables data scientists to effortlessly query feature data stored in Lakehouse tables without requiring SQL expertise. The Python API provides method calls to create point-in-time consistent training data (without future data leakage) from source tables.

It removes the need for users to write complex temporal queries when working with time-series data, as Python API calls are transparently transpiled into SQL, executing everything from push-down filters to ASOF LEFT JOINs.

HQS is a query engine for AI data, powered by Arrow and DuckDB, that provides class-leading performance when working with Lakehouse data. Hopsworks natively supports Arrow data end-to-end, eliminating the bottlenecks caused by serialization and data transposition.

This results in significantly faster data processing, with benchmarks showing an order of magnitude improvement in performance, making data scientists more productive and allowing them to iterate more rapidly.

What if your Lakehouse tables are located on different data warehouses? HQS supports federated queries, enabling data scientists to join and query data across multiple backend data warehouses, such as Snowflake and BigQuery, without duplicating data.

This flexibility is particularly valuable in environments with diverse data storage needs, enabling efficient use of existing resources while maintaining high query performance.