Stage 5: Exploration of AI Application Scenarios

A 5-day intensive AI Agents course co-launched by Google and Kaggle, teaching how to build, evaluate, and deploy production-grade AI agent systems.

Start Learning AI →
AIAgentsGoogleGeminiWebSiteVideoFreeEnglish

5-Day AI Agents Intensive Course – Google & Kaggle

Course Overview

This is a free online course meticulously designed by Google’s machine learning researchers and engineers to help developers master the skills needed to build, evaluate, and deploy AI Agents. It serves as an advanced follow-up to the previously successful Generative AI Intensive course co-hosted by Google and Kaggle, which attracted over 280,000 learners.

Basic Information

  • Course Dates: November 10–14, 2025 (5 days)
  • Format: Fully online
  • Cost: Completely free
  • Organizers: Google & Kaggle
  • Language: English
  • Expected Participants: Hundreds of thousands of learners

Course Highlights

  1. Theory Meets Practice: Each day includes in-depth conceptual explanations, hands-on labs (Codelabs), and live discussions.
  2. Diverse Learning Materials:
    • AI-generated podcasts (created with NotebookLM)
    • Expert whitepapers (authored by Google specialists)
    • Practical coding labs
    • YouTube livestream sessions
    • Discord community support
  3. Capstone Project: A final project after the course; outstanding submissions may win prizes and be featured on Google and Kaggle’s official social media channels.
  4. Completion Badge: Earn a Kaggle badge upon successfully completing the Capstone project.

Course Structure & Tech Stack

Core Technologies

  • Foundation Model: Google Gemini
  • Development Framework: Agent Development Kit (ADK)
  • Communication Protocol: Model Context Protocol (MCP)
  • Platform: Kaggle (for Codelabs)
  • Programming Language: Python
  • Auxiliary Tool: NotebookLM

Core Components

The course covers five essential components of AI Agents:

  1. Models: Foundation large language models
  2. Tools: External functions and API calls
  3. Orchestration: Workflow coordination
  4. Memory: Short-term and long-term memory management
  5. Evaluation: Quality assurance and performance metrics

Daily Breakdown

Day 1: Introduction to AI Agents & Architecture

Theme: Understanding the distinction between Agent systems and LLMs

Key Topics:

  • Agent capability taxonomy
  • Necessity of "Agent Ops" standards
  • Importance of interoperability and security
  • Building reasoning Agents with ADK
  • Orchestrating the Think-Act-Observe loop
  • Exploring multi-Agent design patterns

Learning Materials:

  • Whitepaper: Introduction to Agents
  • Podcast: Unit 1 Summary (via NotebookLM)
  • Codelabs:
    • Building Agentic Apps with the Agent Development Kit (ADK)
    • Introduction to LangGraph
    • Function Calling with the Gemini API

Skills Gained:

  • Understand fundamental Agent architecture
  • Master the Agent loop mechanism
  • Learn to build basic Agents using ADK

Day 2: Agent Tools & MCP Interoperability

Theme: Performing real-world actions via APIs and tools

Key Topics:

  • Integration of external tools and functions
  • Real-time data retrieval
  • Introduction to Model Context Protocol (MCP)
  • Handling complex, long-running operations

Learning Materials:

  • Whitepaper: Agent Tools & Interoperability with Model Context Protocol (MCP)
  • Podcast: Unit 2 Summary (via NotebookLM)
  • Codelabs:
    • Function Calling with the Gemini API
    • Using tools with LangGraph
    • Prompt Caching with the Gemini API

Skills Gained:

  • Implement Function Calling
  • Use the MCP protocol
  • Integrate external tools and APIs

Day 3: Context Engineering & Memory

Theme: Implementing short-term and long-term memory systems

Key Topics:

  • Context engineering practices
  • Dynamic information assembly and management
  • Sessions – immediate interaction history
  • Memory – long-term persistence
  • Creating stateful, personalized AI experiences

Learning Materials:

  • Whitepaper: Context Engineering: Sessions & Memory
  • Podcast: Unit 3 Summary (via NotebookLM)
  • Codelabs:
    • Context Caching with the Gemini API
    • Long-term Memory using LangGraph
    • Personalization with Gemini API

Skills Gained:

  • Manage context windows effectively
  • Build memory systems
  • Create personalized Agent experiences

Day 4: Quality, Logging & Evaluation

Theme: Observability, tracing, and performance metrics

Key Topics:

  • Comprehensive evaluation framework for Agent quality assurance
  • Fundamentals of observability (Logs, Traces, Metrics)
  • Scalable feedback loops
  • LLM-as-a-Judge methodology
  • Human-in-the-Loop (HITL) workflows

Learning Materials:

  • Whitepaper: Agent Quality
  • Podcast: Unit 4 Summary (via NotebookLM)
  • Codelabs:
    • LangSmith Fundamentals
    • Evaluating Agentic Applications
    • A/B Testing with Agents

Skills Gained:

  • Implement quality evaluation systems
  • Use logging and tracing tools
  • Design effective evaluation metrics

Day 5: From Prototype to Production

Theme: Deploying and scaling multi-Agent systems

Key Topics:

  • Agent operational lifecycle (deployment, scaling, productionization)
  • Transitioning from prototype to enterprise-grade solutions
  • Agent2Agent (A2A) protocol
  • Deployment on Vertex AI Agent Engine
  • Multi-Agent system architecture

Learning Materials:

  • Whitepaper: Prototype to Production
  • Podcast: Unit 5 Summary (via NotebookLM)
  • Codelabs:
    • Multi-agent Workflows with LangGraph
    • Human-in-the-loop Workflows in LangGraph
    • Deploying Agents with Google Cloud

Skills Gained:

  • Deploy production-ready Agents
  • Build collaborative multi-Agent systems
  • Scale using Google Cloud infrastructure

Capstone Project

Timeline

  • Launch Date: November 14, 2025 (after Day 5)
  • Submission Deadline: November 30, 2025, 11:59 PM PT
  • Development Window: Approximately 2 weeks

Requirements

  1. Design and implement an AI Agent using the Agent Development Kit (ADK)
  2. Demonstrate at least three core capabilities covered in the course (e.g., tool usage, memory, evaluation)
  3. Provide a Kaggle notebook documentation
  4. Include a short demo video and project description

Rewards

  • All Completers: Receive a Kaggle profile badge
  • Top 10 Submissions:
    • Exclusive Kaggle merchandise
    • Public feature on official Google and Kaggle social media accounts

Time Commitment

  • Daily Assignments: 1–2 hours
  • Live Sessions: 45–60 minutes (optional; recordings available)
  • Total Daily Time: ~2–3 hours
  • Flexibility: Self-paced learning suitable for full-time professionals or students

Target Audience

Ideal For

✅ Developers with prior AI/ML/LLM experience looking to advance into Agent development
✅ Professionals working on automation workflows or multi-step tasks
✅ Learners willing to dedicate time to hands-on labs and live sessions
✅ Individuals aiming to build tangible projects for organizational use or portfolio
✅ Data scientists, ML engineers, and software developers

Less Suitable For

❌ Absolute beginners with no AI/ML background still learning basic LLM usage
❌ Those unable to commit 1–2 hours per day

Prerequisites

  • Basic Python programming (helpful but not mandatory)
  • Interest in AI, machine learning, or data science
  • Familiarity with Python notebooks and common Agent libraries

Learning Outcomes

Upon completion, you will be able to:

  1. ✅ Design a clear Agent framework (Goal → Plan → Tool Call → Memory → Evaluation)
  2. ✅ Gain hands-on experience through Codelabs and the Capstone project
  3. ✅ Understand production considerations (logging, fault tolerance, scalability)
  4. ✅ Earn a completion badge and join a professional community
  5. ✅ Build systems ranging from simple Agents to complex multi-Agent architectures
  6. ✅ Master Agent evaluation and optimization techniques

Why Learn AI Agents?

According to Google: Enterprises are shifting from isolated model calls to Agent platforms that enable scalable experimentation and deployment. Agents have become the way teams transform powerful models into functional software—by wrapping LLMs with memory, tool use, planning, and evaluation to reliably execute multi-step tasks.

Agent vs. LLM

  • Traditional LLM: Single prompt-response interaction
  • AI Agents:
    • Can plan
    • Call tools
    • Coordinate with other Agents
    • Possess memory
    • Evaluate and optimize performance
    • Reliably complete multi-step tasks

Course Resources

Official Links

Learning Tips

  1. Prepare Your Environment Early: Ensure familiarity with Python notebooks and common Agent libraries.
  2. Start with a Real Problem: Bring a small automatable workflow (e.g., weekly reports, support ticket classification, content QA).
  3. Define Evaluation Metrics Early: Draft simple success criteria (task completion rate, latency, reduced manual review time).
  4. Days 1–2: Master fundamentals—Agent loop, tool calling, memory.
  5. Day 3: Add evaluation early—even basic checks prevent misaligned optimization.
  6. Day 4: Only adopt multi-Agent or multi-tool setups if your use case demands it.

Competitive Advantage

As one of the earliest structured courses on AI Agents, completing this program will give you:

  • 🎯 First-mover advantage while many organizations are still figuring out Agent adoption
  • 🎯 A concrete, showcase-ready project for stakeholders or your portfolio
  • 🎯 Access to professional networks within Google and Kaggle communities
  • 🎯 A leap from “understanding Agents” to “delivering Agents”

Community & Support

  • Discord Channel: Actively maintained and supported by Google staff
  • Live Expert Sessions: Q&A with Google researchers and engineers
  • Global Learning Community: Hundreds of thousands of peers learning together
  • Session Recordings: All livestreams recorded for learners in different time zones

Summary

The 5-Day AI Agents Intensive Course offers a timely and rigorous opportunity for developers and ML practitioners to transition from “LLM applications” to “Agent systems.” As AI evolves from offering suggestions to taking actions, those who understand how to build Agents with memory, tools, coordination, evaluation, and deployment capabilities will hold a significant edge.

This is more than just a course—it’s your passport into the era of AI Agents.