Google Five Day AI Agent Training Materials All Playlist https://www.youtube.com/playlist?list=PLqFaTIg4myu9r7uRoNfbJhHUbLp-1t1YE
Day 1
🎒 Today’s Assignments
Complete the Unit 1 – “Introduction to Agents”:
Listen to the summary podcast episode for this unit. To complement the podcast, read the “Introduction to Agents” whitepaper. Complete these codelabs on Kaggle: Build your first agent using Gemini and Agent Development Kit (ADK). Build your first multi-agent systems using ADK. Make sure you phone verify your Kaggle account before starting, it's necessary for the codelabs. We also have a troubleshooting guide for the codelabs. Be sure to check there for solutions to common problems. Want to have an interactive conversation? Try adding the whitepaper to NotebookLM. 💡 What You’ll Learn
Today’s whitepaper introduces Al agents. It presents a taxonomy of agent capabilities, emphasizes the need for an Agent Ops discipline for reliability and governance, and discusses the importance of agent interoperability and security through identity and constrained policies.
In today's codelabs, you'll be building your first AI agent and your first multi-agent system, using ADK, powered by Gemini, and giving it the ability to use Google Search to answer questions with up-to-date information. In the second codelab, the focus will be on multi-agent systems, where you'll learn how to create teams of specialized agents and explore different architectural patterns.
đź“‹ Reminders
Tomorrow at 11:00 AM PT, Kanchana Patlolla and Anant Nawalgaria will host the first livestream on Kaggle’s YouTube channel. They’ll be joined by codelab authors Kristopher Overholt and Hangfei Lin, along with other special guests from Google: Alan Blount, Mike Clark, Michael Gerstenhaber and Antonio Gulli to discuss the assignments and share insights. This course page serves as the central hub for all event resources, including assignments and important updates. Kaggle's Discord is the best place to ask questions. In addition to other participants, several Googlers are there to help. During the livestream, we'll also pick some questions from Discord to discuss. You'll get Kaggle swag if your question is chosen! We want this community to be positive and supportive. Please follow Kaggle’s community guidelines found here.
Day 2
Complete Unit 2 - “Agent Tools & Interoperability with Model Context Protocol (MCP)”:
Listen to the summary podcast episode for this unit, created by NotebookLM. To complement the podcast, read the “Agent Tools & Interoperability with MCP” whitepaper. Complete these codelabs on Kaggle: Explore new ways to add tools to extend what your agents can do. Explore best practices for tools, including using MCP and long-running operations. Make sure you phone verify your Kaggle account before starting, it's necessary for the codelabs. We also have a troubleshooting guide for the codelabs. Be sure to check there for solutions to common problems. Want to have an interactive conversation? Try adding the whitepaper to NotebookLM.
Day 3 Complete Unit 3 - “Context Engineering: Sessions & Memory”:
Listen to the summary podcast episode for this unit, created by NotebookLM.
To complement the podcast, read the “Context Engineering: Sessions & Memory whitepaper".
Complete these codelabs on Kaggle:
Build stateful agents and perform context engineering.
Explore how to use memory with your agent.
We also have a troubleshooting guide for the codelabs. Be sure to check there for solutions to common problems.
Want to have an interactive conversation? Try adding the whitepapers to NotebookLM.
💡 What You’ll Learn
This whitepaper explores context engineering as the practice of dynamically assembling and managing information within an agent's context window to create stateful and personalized Al experiences. It defines Sessions as the container for a single, immediate conversation's history, and Memory as the long-term persistence mechanism.
Day 4 Complete Unit 4 - “Agent Quality”:
Listen to the summary podcast episode for this unit, created by NotebookLM. To complement the podcast, read the Agent Quality whitepaper. Complete these codelabs on Kaggle: Implement observability to help you debug your agents. Evaluate your agents. Be sure to check the troubleshooting guide solutions to common problems. Want to have an interactive conversation? Try adding the whitepapers to NotebookLM. 💡 What You’ll Learn
This whitepaper addresses the challenge of assuring quality in Al agents by introducing a holistic evaluation framework. The necessary technical foundation for this is Observability, built on three pillars: Logs (the diary), Traces (the narrative), and Metrics (the health report), enabling a continuous feedback loop using scalable methods like LLM-as-a-Judge and Human-in-the-Loop (HITL) evaluation.
Day 5 Complete Unit 5 - “Prototype to Production”:
Listen to the summary podcast episode for this unit. To complement the podcast, read the “Prototype to Production” whitepaper. Complete these codelabs on Kaggle: Explore how to use Agent2Agent (A2A) Protocol to have agents interact with each other. [Optional] Deploy your agent to Agent Engine on Google Cloud. 💡 What You’ll Learn
This whitepaper provides a technical guide to the operational lifecycle of AI agents, focusing on deployment, scaling and productionization. It explores the challenges of transitioning agentic systems from prototypes to enterprise-grade solutions, with special attention to A2A Protocol.
For today's codelabs, you'll learn how to build systems of multiple, independent agents that can communicate and collaborate using A2A Protocol. You'll also learn how to take your agent from your local machine to a production-ready, scalable service, by deploying your agent to Vertex AI Agent Engine on Google Cloud.
For today's codelabs, you'll learn how to use logs, traces, and metrics to get full visibility into your agent's decision-making process, allowing you to debug failures and understand why your agent behaves the way it does. In the second codelab, you'll learn how to evaluate your agents to score your agent's response quality and tool usage.