Prototype to Production with ADK

Опубликовано 30 марта 2026, 19:00

Learn more about the Agent Development Kit → google.github.io/adk-docs

You've seen impressive AI agent demos, but what is the playbook for taking a simple proof-of-concept to an enterprise-grade agent that your business can actually rely on? In this video, we'll show you how to build, deploy, and monitor a robust, scalable, and secure multi-agent application from scratch using Google's Agent Development Kit (ADK).

Join Sita and Ayo as they walk you through building a complete AI Code Review Assistant. You will learn how to orchestrate specialized multi-agent systems, manage shared context with session states and persistent memory, deploy your agent as a scalable service to Vertex AI Agent Engine, and debug execution using Cloud Trace.

Chapters:
0:00 - Introduction
1:38 - Code Review Assistant Architecture
3:12 - ADK Concepts: Sessions, Memory, & Artifacts
4:56 - Agent Deployment Options
5:49 - Observability with Cloud Trace
6:29 - Chapter 1: Intro & Learning Objectives
6:49 - Chapter 2: Your First Agent Deployment
11:18 - Chapter 3: Preparing Your Workspace
14:26 - Chapter 4: Building Your First Agent
18:38 - Chapter 5: Sequential Agent Pipeline (Review)
28:36 - Chapter 6: Loop Agent Pipeline (Fix)
37:24 - Chapter 7: Deploying to Production
41:49 - Chapter 8: Observability
42:55 - Eng Talk: Community questions
44:46 - Outro & next steps

Resources:
Codelab → goo.gle/adk-code-reviewer-assi...
ADK Samples GitHub repo → goo.gle/4uVtmSR
More videos on ADK → goo.gle/4uUiEfp

#AIAgents #ADK #GoogleCloud #VertexAI #MultiAgentSystems #ProductionAI #observability

Subscribe to Google For Developers → goo.gle/developers

Products Mentioned: ADK, Agent Development Kit, Google AI