AI Agents Intensive — 5-Day Course

This repository contains notes and exercises from the 5-day "AI Agents Intensive" (hosted via Kaggle/Google). The course covers practical foundations for building production-ready AI agents, including models, tools, orchestration, memory, and evaluation.

Badges: (add CI / license / coverage badges here)

Table of Contents

• Overview
• Daily Modules & Assignments
  • Day 1 — Introduction to Agents
  • Day 2 — Agent Tools & MCP
  • Day 3 — Context Engineering: Sessions & Memory
  • Day 4 — Agent Quality
  • Day 5 — Prototype to Production
• Source

Overview

This 5-day program explores how agents go beyond simple LLM prototypes and become robust, reliable systems suitable for production. Key focus areas include:

• Models — capabilities and selection
• Tools — safe external actions and tool design
• Orchestration — multi-agent communication and workflows
• Memory — short-term sessions and long-term persistence
• Evaluation — observability, metrics, and human-in-the-loop (HITL)

Daily Modules & Assignments

Day 1 — Introduction to Agents

• Whitepaper highlights
• Taxonomy of agent capabilities
• Agent Ops: reliability and governance
• Identity, policies, and secure interoperability

Codelabs / Exercises:

• Build your first AI agent using Gemini + ADK
• Build a multi-agent system
• Use Google Search as an external tool

Day 2 — Agent Tools & MCP (Model Context Protocol)

• Whitepaper highlights
• External tool functions for real-time actions
• Best practices for tool design and safety
• MCP introduction: communication patterns, risk model, readiness signals

Codelabs / Exercises:

• Turn Python functions into agent-executable tools
• Use MCP for interoperability
• Implement long-running operations with human approval

Day 3 — Context Engineering: Sessions & Memory

• Whitepaper highlights
• Context engineering for stateful, personalized agents
• Sessions: short-term conversational state
• Memory: persistent, long-term storage and retrieval

Codelabs / Exercises:

• Build stateful agents using conversation history
• Use working memory inside a session
• Implement long-term memory across sessions

Day 4 — Agent Quality

• Whitepaper highlights
• Holistic evaluation framework for agents
• Observability foundations: logs, traces, metrics
• Techniques: LLM-as-Judge, human-in-the-loop (HITL)

Day 5 — Prototype to Production

• Whitepaper highlights
• Deployment, scaling, and productionization patterns
• A2A Protocol for multi-agent communication
• Transitioning from prototype to enterprise-grade systems

Codelabs / Exercises:

• Build multi-agent systems using A2A
• (Optional) Deploy agents to Vertex AI Agent Engine

Source

Materials and inspiration from: https://www.kaggle.com/learn-guide/5-day-agents

---

Repository: aiagent_course — notes and code for the 5-day intensive.