Snowflake, the AI Data Cloud company, today announced new AI features to its platform, enabling developers to effortlessly build conversational apps for structured and unstructured data with high accuracy, efficiently batch inference of large language models (LLM) for natural language processing (NLP) pipelines, and train custom models with GPU-powered containers — all with built-in governance, access controls, observability, and safety guardrails ensuring that AI security and trust remain at the forefront. The launches make it easier for companies to deliver simple, effective and reliable AI solutions for all data, including structured and unstructured data.
Snowflake is also introducing Snowflake Intelligence (available soon in private preview), a new platform where companies can ask questions of their business data, and based on the results, quickly create data agents that take action based on these insights, as well as advancements in Cortex AI that empower companies to develop high-quality conversational applications. This and much more was presented during BUILD 2024 – Snowflake’s annual event for developers.
Snowflake enables companies to unlock value across data silos
Enterprise data is often spread out in silos on different storage platforms, making it difficult to use to create value. To improve the conditions for companies to create value from their data, Snowflake has launched the Snowflake Open Catalog, now publicly available. It allows companies to easily integrate and secure data within open lakehouse structures, such as Apache Iceberg based systems. The service ensures consistent governance and management across different engines, providing flexibility, while maintaining reliability, security and scalability.
To help businesses scale out unnecessary expenses, improve performance, and transform data lakes into more open lakehouse structures with Apache Iceberg, a range of new streaming, data ingestion, change data capture pipelines, and integrations are also being launched.
Snowflake’s service for analyzing unstructured data, Document AI, is now available on AWS and Microsoft Azure. Powered by Snowflake’s advanced Arctic-TILT model, Document AI extracts valuable insights from text-heavy information and other elements in documents, such as logos and handwritten text. The service has already been adopted by several large organizations to analyze text-heavy documents for valuable insights using NLP.
The future of Data Agents for Enterprise AI
Snowflake Intelligence, soon available in private preview, will allow users to ask natural language questions of their business data to enhance understanding, and generate actionable insights.
Snowflake Intelligence will provide access to best-in-class data agents built on Snowflake’s data foundation, while protecting corporate IP and delivering answers based on credible corporate data. In addition, Snowflake Intelligence makes it easy to access and utilize the full potential of both on-premises data and that found in various third-party tools such as SharePoint, Slack, Salesforce, and Google Workspace.
Build high-quality conversational apps – faster
Thousands of companies worldwide rely on Cortex AI to scale and deploy their AI applications. Snowflake’s latest releases for the platform make it easier for companies to build trusted AI apps with both structured and unstructured data sources, easier orchestration, and built-in evaluation and monitoring. Here are some of the key launches for Snowflake Cortex AI.
Create more engaging responses with multimodal support
AI applications can now be enhanced by multimodal inputs such as images, and soon audio and other data types, using LLM models like Meta’s Llama 3.2 and the new Cortex COMPLETE Multimodal Input Support (soon available for private preview).
Comprehensive answers with new connections to knowledge bases
Users can quickly integrate with internal knowledge bases using the new Snowflake Connector for SharePoint (now available for public preview), allowing direct integration with Microsoft 365 SharePoint files. Cortex Knowledge Extensions in the Snowflake Marketplace (soon available for private preview) also enable seamless usage of third-party data, such as news articles and scientific publications, with copyright protection and clear attribution to the author.
Increased reliability and improved compliance in applications with built-in evaluation and monitoring
Generative AI apps can now be evaluated and monitored using over 20 methods for relevance, anchoring, stereotyping and response times during both development and operation. This with AI Observability for LLM apps (now available as a private preview) integrated with TruEra technology.
Reduce manual integration and coordination work
To simplify the management of enterprise data queries, developers can use the Cortex Chat API (soon available for public preview) to streamline integration between application front ends and Snowflake. The Cortex Chat API combines structured and unstructured data into a single REST API call, helping developers quickly create Retrieval-Augmented Generation (RAG) and analytic AI agents with less effort.