Introducing Feature Views
Databricks, Friday, July 10th, 2026
Databricks launches Feature Views in public preview, letting ML teams define features once and deploy everywhere.
Databricks announced the public preview of Feature Views, a managed framework that lets machine learning teams define features once and deploy them across experimentation, training, and production. The system automatically handles data pipelines, eliminates training-serving skew, and supports both batch and real-time streaming sources with end-to-end latency of 200 milliseconds.
Features become governed Unity Catalog objects, reducing operational complexity around online stores and infrastructure while enabling data scientists to move from feature conception to production deployment rapidly.