AI Agentic Boot Camp

Build Production-Ready AI Agents from Scratch

Welcome to the AI Agentic Boot Camp

This 7-week bootcamp teaches you how to build AI agents that can use tools, connect to external services, and solve real problems. You’ll learn practical development skills through hands-on projects.

What You’ll Build

By the end of this bootcamp, you’ll have hands-on experience creating:

  • Custom AI Agents using LangChain and LangGraph
  • Tool-Enabled Systems that interact with external APIs and services
  • Production-Ready Applications with proper observability and scaling
  • MCP-Connected Agents that leverage the Model Context Protocol

Who This Is For

This bootcamp is for anyone interested in learning about agentic AI. It was especially created for the BYU-I Data Science Society, but all are welcome.

Ideal participants:

  • Have basic Python programming experience
  • Want to build practical AI applications (not just theory)
  • Are ready to dive deep into modern agent architectures
  • Understand that AI development is about engineering, not just prompts

Course Overview

Learning Path

This bootcamp follows a structured progression from fundamentals to production deployment:

Weeks 1-3: Foundations
Master the core concepts of LLMs, agents vs. workflows, and tool integration through the Model Context Protocol.

Weeks 4-5: Professional Development
Learn production-grade practices including observability with LangSmith, debugging complex agents, and advanced topics.

Weeks 6-7: Production & Beyond
Scale your agents for real users, manage context and costs, understand model maintenance, and prepare for long-term deployment.

Technologies You’ll Master

  • LangChain - Framework for building LLM applications
  • LangGraph - Stateful multi-step agent workflows
  • LangSmith - Observability, tracing, and evaluation
  • Model Context Protocol (MCP) - Standardized tool and data connections
  • Google Gemini API - Primary LLM provider (OpenAI compatible)
  • Production Tools - Memory management, cost optimization, scaling strategies

Key Learning Outcomes

By completing this bootcamp, you will be able to:

✓ Design and implement custom AI agents with tool-calling capabilities
✓ Build workflows that balance predictability and autonomy
✓ Connect agents to external services using MCP servers
✓ Debug and optimize agent behavior with LangSmith tracing
✓ Deploy production-ready agents that scale to multiple users
✓ Manage costs, context, and performance in real-world scenarios
✓ Understand when to retrain models and maintain AI systems over time

Quick Start Guide

Prerequisites

Before starting the bootcamp, make sure you have:

  1. Google Account - Required for Google Colab and Gemini API access
  2. Basic Python knowledge - Familiarity with functions, variables, and imports
  3. Web browser - Chrome or Firefox recommended for Colab
  4. Internet connection - For accessing Colab notebooks and APIs

No local installation required! All coding happens in Google Colab, which provides a free cloud environment with Python pre-installed.

Getting Your API Key

The bootcamp uses Google’s Gemini API (free tier available):

  1. Visit Google AI Studio
  2. Sign in with a personal Google account (not school email)
  3. Generate an API key
  4. Store it securely (we’ll use .env files in the course)

Alternative: If Gemini isn’t available in your region, OpenAI’s API works with minor code adjustments ($5 minimum deposit required).

Working Environment

All coding is done in Google Colab notebooks:

  • No local setup required
  • Free GPU access
  • Automatic dependency installation
  • Easy sharing and collaboration

Access all notebooks on the Notebooks → page.

Your First Steps

  1. Week 1: Get your Gemini API key
  2. Module 1: Open the first Colab notebook
  3. Save a copy to your Google Drive
  4. Follow along with the lesson structure
  5. Experiment with code examples

Weekly Breakdown

Week 1 - LLM Bootcamp

Focus: Getting Started with AI Agents

  • Create custom chats using API calls
  • Secure and use Gemini API keys
  • Understand LLM concepts and services
  • Explore LangChain basics
  • Learn prompt and context engineering

View Module 1 →

Week 2 - Workflows and Agents

Focus: Understanding LLM Architecture Patterns

  • Distinguish between workflows and agents
  • Learn when to use each approach
  • Plan and diagram agent behaviors
  • Translate conceptual workflows into code
  • Explore no-code/low-code options (n8n, Langflow)

View Module 2 →

Week 3 - Tools and MCP Connections

Focus: Connecting AI Agents to External Services

  • Understand the Model Context Protocol (MCP)
  • Set up Canvas MCP connections
  • Build agents that use tools
  • Implement User → AI → MCP → Tool flow
  • Create real-world integrations

View Module 3 → | Open in Colab →

Week 4 - LangSmith Diagnostics

Focus: Observability and Debugging

  • Use LangSmith for agent observability
  • Trace execution and identify bottlenecks
  • Debug tool calls and agent decisions
  • Build evaluation datasets
  • Analyze trace trees for optimization

View Module 4 → | Wellness Agent → | Canvas Tutor →

Week 5 - Advanced Topics

View Module 5 →

Week 6 - Production & Scaling

Focus: From Prototype to Production

  • Understand production vs. development environments
  • Manage token costs at scale
  • Build multi-user stateful interfaces
  • Prevent context overflow and runaway costs
  • Design resilient, scalable architectures

View Module 6 → | Open in Colab →

Week 7 - Retraining and Weights

Focus: Model Maintenance and Drift

  • Understand data drift and model degradation
  • Detect when retraining is necessary
  • Update model weights responsibly
  • Build authentic, maintainable AI systems
  • Plan long-term model lifecycle

View Module 7 → | Open in Colab →

Additional Resources

Course Materials

  • Notebooks - All Google Colab notebooks with project descriptions
  • Reference - Documentation links and key terms dictionary

Getting Help

  • Review error messages carefully—they often point directly to the solution
  • Check the Reference → page for unfamiliar terms
  • Revisit earlier modules if concepts feel unclear
  • Experiment with code variations to deepen understanding

Ready to Begin?

Start with Module 1 - LLM Bootcamp → and work through each week sequentially. The course is designed to build progressively—each module assumes knowledge from previous weeks.

Let’s build something intelligent.