Key Points
- Five New Tools: Databricks launches Mosaic AI Agent Framework, Agent Evaluation, Tools Catalog, Model Training, and Gateway.
- Improved Reliability: Focus on quality, cost-efficiency, and data privacy.
- Modular Systems: Enhanced control for mission-critical AI applications.
- Enterprise Integration: Unified interface for managing and deploying AI models.
Expanding the Mosaic AI Suite
At the Data + AI Summit, Databricks announced significant expansions to its Mosaic AI platform, originally developed from the $1.3 billion acquisition of MosaicML. The company introduced five new tools designed to enhance enterprise capabilities in building and deploying large language models (LLMs).
🤖🇺🇸 Databricks revolutionizes enterprise AI with new Mosaic AI tools! Discover how this powerhouse empowers businesses to build with advanced large language models (LLMs). Exciting times for AI enthusiasts! #AI #LLMs pic.twitter.com/ttj7FSopMh
— Talk AI Today (@TalkAIToday) June 12, 2024
Key New Features
- Mosaic AI Agent Framework: This tool empowers developers to build Retrieval Augmented Generation (RAG)-based applications by leveraging Databricks’ serverless vector search functionality. It offers a hybrid approach combining classic keyword-based search with embedding search, integrated with Databricks’ data lake.
- Mosaic AI Agent Evaluation: An AI-assisted evaluation tool that uses LLM-based judges for production testing and user feedback collection. It integrates a UI component from Databricks’ acquisition of Lilac, aiding in visualizing and searching large text datasets.
- Mosaic AI Tools Catalog: Provides a centralized repository for AI tools and functions, enhancing discoverability and governance. This catalog allows enterprises to control which AI tools can be used, ensuring compliance and security.
- Mosaic AI Model Training: Enables enterprises to fine-tune models with proprietary data, improving performance on specific tasks. This service caters to the need for customized AI solutions tailored to unique organizational requirements.
- Mosaic AI Gateway: A unified interface for querying, managing, and deploying various AI models. It offers centralized credential storage, rate limits for vendors, usage tracking, and debugging capabilities, ensuring secure and cost-effective AI model deployment.
Enhancing AI with Modular Systems
Databricks’ co-founders, CEO Ali Ghodsi and CTO Matei Zaharia, emphasized the shift towards modular AI systems. These systems allow enterprises to control all aspects of AI applications by chaining together multiple models and functions. This approach enhances reliability, cost-efficiency, and data privacy, addressing core concerns of AI deployment.
Industry Insights
Ali Ghodsi highlighted the evolving landscape of AI, where enterprises increasingly prioritize open models and sophisticated tools. “The big shift happened in the last quarter and a half. Customers now adopt an entirely new set of tools to tackle problems and opportunities with open models,” said Ghodsi.
Matei Zaharia added, “High-impact, mission-critical AI applications require modular systems. Developers can control all aspects of AI workflows, ensuring trustworthiness and relevance.”
Real-World Applications and Governance
The new Mosaic AI tools align with Databricks’ vision of supporting enterprises in their AI journey. The AI Agent Framework and Tools Catalog, for example, enable developers to create advanced RAG-based applications, while the AI Gateway ensures secure and governed model usage.
By integrating these tools with Databricks’ Unity Catalog governance layer, enterprises can maintain strict control over personal data and AI functionalities. This comprehensive approach ensures that AI deployments are not only effective but also compliant with data privacy regulations.
Databricks’ launch of the new Mosaic AI tools marks a significant step forward in enterprise AI solutions. By focusing on quality, cost-efficiency, and data privacy, Databricks is equipping enterprises with the tools they need to build reliable, high-impact AI applications. As the AI landscape continues to evolve, these innovations will undoubtedly play a crucial role in shaping the future of enterprise AI.
