Vector database and more flexibility: Databricks optimizes platform for LLMs

The provider Databricks has announced that the Vector Search function for its data intelligence platform is now available without restrictions. Vector Search is a serverless vector database specially developed for enterprise use. It makes it possible to extend Large Language Models (LLMs) with your own company data. In addition, the company founded by the creators of Apache Spark has announced comprehensive additions for model serving in the context of Retrieval Augmented Generation (RAG). For example, the Foundation Model API in Model Serving gives companies the option of using and querying LLMs via a serving endpoint.

More flexibility and quality for RAG applications

The innovations presented are intended to help simplify the construction and productive use of applications with generative AI and ensure a high standard of quality. Among other things, users will be supported by a redesigned user interface and a more flexible quality control interface for monitoring RAG applications in productive operation. In addition to the new vector search, additional language models such as Claude3, Gemini, DBRX and Llama3 can also be integrated into the data intelligence platform. Databricks also promises higher performance when providing and querying LLMs, as well as improvements in terms of governance and auditability of the applications.

data2day 2024: From data mesh to GenAI - a new era in data science

How are LLMs, GenAI and knowledge graphs changing data science and which data architecture is proving its worth in practice? data2day provides answers to these and many other questions. The conference for data scientists, data engineers and data teams will take place in Heidelberg on September 18 and 19, 2024.

Further information and details on Vector Search, Model Serving and the Foundation Model API can be found in the announcement on the Databricks blog.

(map)