3x Faster Search: Parallel Test-Time Scaling with Instructed-Retriever-1
Summary
Instructed-Retriever-1 now delivers 3x faster search for Agent Bricks Knowledge Assistant. This parallel test-time scaling update also improves quality for Databricks practitioners.
Summary generated by brickster.ai. For the full article, follow the source link above.
More from Databricks Blog
From Wall Street to Data Platforms
Databricks values deep industry expertise, as shown by Kim Hatton’s transition from finance to helping financial institutions solve modern data challenges. Our collaborative environment encourages employees to grow beyond their core roles and contribute to industry innovation, building practical tools that turn complex data tasks into streamlined successes.
Enabling Evolutionary Database Development: Database branching with Lakebase, the conclusion
Lakebase now supports database branching, enabling evolutionary database development. This concludes the series on Lakebase's operationalization of evolutionary database design.
What is customer segmentation?
Customer segmentation combines multiple types and methods, from rule-based to AI/ML-driven models, but its success hinges on unifying fragmented customer data into a governed Customer 360. Databricks' CustomerLake, an Agentic CDP, builds segments directly on governed data with AI-driven identity resolution and natural-language audience creation, eliminating data copies and extra vendors.
Unlocking semantics for AI: How Mercedes-Benz Korea built trusted “Talk to Data” at scale
Mercedes-Benz Korea built a trusted "Talk to Data" solution at scale by making 500+ KPI definitions available in an AI-ready semantic layer on Unity Catalog metric views, accelerating the transition with an automated DAX-to-Metric-View transpiler. This governed semantic layer supports both existing BI and new "Talk to Data" experiences, with Genie and Agent Bricks providing consistent answers and shaping a playbook for persona-based AI agents across markets.
Forward Deployed Engineering: Delivering Business Outcomes with AI
Databricks is launching its Forward Deployed Engineering (FDE) organization to accelerate customer business outcomes with AI, pairing the Lakehouse platform with embedded, engineering-led delivery. This new approach moves beyond migration and pipeline building to solve business problems with production AI agents, as demonstrated by customers like Fox, JPMC, and Qualcomm.
Ingesting the Milky Way: Petabyte-Scale with Zerobus Ingest
Zerobus Ingest, a new serverless streaming API, enables instant deployment of petabyte-scale data pipelines on Databricks without manual infrastructure management. Its dynamic partitioning architecture automatically scales compute and sustains over 12 GB/s throughput to a single table, efficiently handling unpredictable data volumes.