At the 2025 Data + AI Summit, Databricks introduced Agent Bricks, a major leap forward in enterprise AI tooling that empowers companies to build and deploy powerful, cost-effective AI agents directly on their proprietary data—without the complexity of manual tuning or piecing together disparate tools.
“It’s a whole new way of building and deploying AI agents that can reason on your data,” said Ali Ghodsi, CEO and co-founder of Databricks. The new solution simplifies the development lifecycle of generative AI agents by integrating all essential capabilities—from data generation and model evaluation to deployment—within the Databricks ecosystem.
Agent Bricks leverages the latest advancements from Mosaic AI’s research arm to generate synthetic, domain-specific training data, which eliminates the need for costly and time-consuming data labeling. The offering also automates rigorous task-specific evaluations to help enterprises select the most accurate and cost-optimized agents for immediate production use.
The platform supports a wide range of applications, including enterprise knowledge assistants, information extraction tools, and custom large language model (LLM) agents, making it relevant across diverse industries such as healthcare, pharmaceuticals, finance, and consumer tech.
Early adopters have already showcased its impact:
- AstraZeneca used Agent Bricks to extract structured data from more than 400,000 clinical trial documents in under an hour—dramatically accelerating time-to-insight in a sector where data parsing often takes weeks or months.
- Flo Health, a women’s health platform, deployed Agent Bricks to enhance its AI-powered medical assistant, focusing on improved accuracy while preserving user privacy. “Agent Bricks delivers higher-quality results at a significantly lower cost,” said Roman Bugaev, Flo Health’s Chief Technology Officer.
In addition to Agent Bricks, Databricks also announced support for serverless GPUs, allowing teams to dynamically scale AI workloads without infrastructure management. This is particularly valuable for running inference and fine-tuning tasks with large models in a more cost-efficient manner.
Databricks further strengthened its end-to-end AI platform with the release of MLflow 3.0, a major upgrade to its open-source machine learning lifecycle management tool. The new version introduces features like prompt management, LLM evaluation, and human-in-the-loop feedback, critical for building and monitoring production-grade generative AI systems. MLflow now boasts over 30 million monthly downloads, cementing its position as one of the most widely used platforms in the machine learning community.
Together, these announcements underscore Databricks’ strategic vision of offering a unified, enterprise-ready stack for generative AI, with simplicity, performance, and security at its core.
As the race to operationalize AI intensifies, Agent Bricks positions Databricks at the forefront—enabling organizations to bring intelligent, custom-built agents to production faster than ever before, using their own data as a strategic asset.
