Description
Vector databases such as ElasticSearch and Pinecone offer fast ingestion and querying on vector embeddings with ANNs. However, they typically do not decouple compute and storage, making them hard to integrate in production data stacks. Because data storage in these databases is expensive and not easily accessible, data teams typically maintain ETL pipelines to offload historical embedding data to blob stores. When that data needs to be queried, they get loaded back into the vector database in another ETL process. This is reminiscent of loading data from OLTP database to cloud storage, then loading said data into an OLAP warehouse for offline analytics. Recently, “lakehouse” offerings allow direct OLAP querying on cloud storage, removing the need for the second ETL step. The same could be done for embedding data. While embedding storage in blob stores cannot satisfy the high TPS requirements in online settings, we argue it’s sufficient for offline analytics use cases like slicing and dicing data based on embedding clusters. Instead of loading the embedding data back into the vector database for offline analytics, we propose direct processing on embeddings stored in Parquet files in…
Description from YouTube. Full content on the video page.
Topics
More from Databricks
NewsHow AI and Data Keep 2.3 Million Lawns Healthy | TruGreen & Databricks
TruGreen uses Databricks Genie and Lakehouse to manage 2.3 million lawns with AI that optimizes service timing and predicts customer churn using weather, soil, and service data. The system enables non-technical branch managers to take daily actions through customized reports without requiring data expertise.
NewsHow FOX Sports Uses AI to Power Search
Fox Sports rebuilt their search system on Databricks to handle rapidly changing sports information by continuously streaming player, team, and content data into the index while computing real-time trends. The system uses semantic vector search with time-weighted ranking to surface fresh content higher, doubling the rate at which users find what they're looking for.
NewsHow AI Helps Match Doctors With Communities in Need | Databricks for Good
Databricks provides pro bono services and compute discounts to help nonprofits stay lean through its Databricks for Good program. The video demonstrates the Virtue Foundation agent, which uses Genie and interactive hex maps to identify underserved populations and medical facility distribution in Kenya.
NewsOmnigent: Open-Source Meta-Harness for AI Agents | Matei Zaharia
Omnigen is an open-source meta-harness developed by Databricks that acts as an orchestration and control layer to wrap, manage, and combine multiple AI coding agents. The platform introduces contextual security policies, cost controls, multi-agent task routing, and sandbox integrations to enable collaborative workflows and centralized governance.
NewsBuilding Agents on Databricks with Custom Apps and Omnigent
This video demonstrates how to build, update, and govern custom AI agents on Databricks using Agent Bricks, Databricks Apps, and Omnigent. The tutorial shows how to integrate Model Context Protocol servers, track execution with MLflow traces, schedule automated agent tasks, and manage security policies through Unity AI Gateway.
NewsGoverning AI Strategy with Unity AI Gateway
The Unity AI gateway provides centralized monitoring, cost tracking, and governance for enterprise AI models, MCP servers, and coding agents. Administrators can use the platform to set budget limits, configure external providers, and enforce input policies that block personally identifiable information.