Summary
Databricks has introduced Feature Views, a managed framework that allows practitioners to define an ML feature once in Unity Catalog and use it consistently across training, batch inference, and real-time serving. This release eliminates training-serving skew and complex infrastructure management, enabling users to rapidly productionize features with a few API calls and serve streaming features at a 200ms end-to-end p99 latency.
Summary generated by brickster.ai. For the full article, follow the source link above.
More from Databricks Blog
Clean data was never enough: How IFS Copperleaf taught AI its business with Databricks Genie One
IFS Copperleaf built a company-wide AI assistant on Databricks Genie One, using Genie Ontology and Unity Catalog Skills to establish shared business logic across nearly 30 disparate systems. By enabling users to query raw data across tools like Salesforce and M365 in plain English, the deployment achieved a 70% activation rate and covered its running costs in four months.
Talk to your clinical data: unlocking AWS HealthLake with OpenSharing and Genie One
A new OpenSharing integration connects AWS HealthLake FHIR data directly to Databricks as a shared catalog, eliminating the need for costly export pipelines and data duplication. Healthcare organizations can now govern clinical data through Unity Catalog, run queries across notebooks and dashboards, and explore insights in natural language with Genie One.
Build a real-time next-best-offer recommender using Databricks Feature Views
A ready-to-run Solution Accelerator allows you to deploy an end-to-end, real-time next-best-offer recommender powered by Databricks Feature Views. The architecture unifies batch and streaming features across training and serving to rank offers against live customer behavior with 34 ms p95 latency.
Teach Genie One your workflows: The ultimate guide for skills in Genie One
Skills in Genie One enable practitioners to operationalize their specific steps, processes, and checks beyond baseline business ontology definitions. Learn how to map your workflows, have Genie One interview you to author the skill, and validate the results against golden references.
Tabular foundation models on Databricks AI Runtime
Tabular foundation models accelerate predictive ML workflows by generating predictions on new datasets without requiring task-specific training. Databricks AI Runtime enables teams to benchmark and fine-tune these models against existing baselines using managed GPUs, MLflow tracking, and Unity Catalog integration.
Introducing the Databricks AI Index: The state of enterprise AI tokenomics
Databricks has launched the Databricks AI Index, a public resource tracking real-world enterprise model adoption, token spending, and coding-agent performance across more than 10,000 organizations using Unity Gateway. The index reveals that multi-provider setups and open-weight models are surging across enterprise workloads, while demonstrating that models with nearly identical coding success rates can vary in cost by over 30x.