Agent Bricks: Build a Multi-Agent GenAI System in Minutes

Abstract

In this hands-on demo, see how Agent Bricks makes it easy to build high-quality, production-ready AI agents—without wrangling infrastructure or writing complex code.

We’ll walk through how to:

• Build a knowledge assistant for an R&D team to query PDFs of energy drink formulas
• Evaluate performance automatically using MLflow traces and custom metrics
• Improve answers with natural language feedback using agent learning—no retraining needed
• Orchestrate expert agents from multiple departments (R&D, marketing, finance) into a multi-agent supervisor that can answer your CEO’s toughest questions, like “What flavor should we launch next?”

You’ll learn how to unify unstructured data, automate evaluation, and streamline agent tuning—all from a simple declarative interface. :chart_with_upwards_trend: From one agent to a full AI system, Agent Bricks lets you build fast, iterate faster, and scale with confidence.

:link: Learn more at databricks.com/product/artificial-intelligence/agent-bricks