Generative AI Foundations

A comprehensive 5-day training program covering the fundamentals of Generative AI, Large Language Models (LLMs), and advanced AI architectures.

Course Overview

This hands-on course takes you from the basics of generative AI to building production-ready agentic systems. Each day builds on the previous, combining theory with practical labs using Azure OpenAI, LangChain, and modern AI frameworks.

---

Day 1: Foundations of Generative AI and LLMs

Topics Covered:

• Evolution of AI to Generative AI
  Understand how AI evolved from rule-based systems to deep learning and transformers, leading to generative models.

• How LLMs Work
  Learn the architecture of transformers and attention mechanisms, and how LLMs are trained to generate coherent text.

• Types of LLMs and Use Cases
  Compare GPT, Claude, Gemini, and LLaMA, and explore their strengths in different domains.

• Grounding and Reliability Techniques
  Discover how grounding improves model relevance and reduces hallucinations, along with techniques like RLHF and tool use.

Lab:

Use Azure OpenAI to analyze delivery feedback and generate summaries. Experiment with different LLMs and observe output differences.

---

Day 2: Prompt Engineering Deep Dive

Topics Covered:

• Types of Prompts
  Learn how basic, structured, and contextual prompts influence model responses.

• Prompt Engineering Techniques
  Explore zero-shot, few-shot, and chain-of-thought prompting to guide model reasoning.

• Prompt Templates
  Design reusable templates for tasks like ETA prediction, delivery issue classification, and customer communication.

• Sampling Parameters
  Understand how temperature and top-p affect creativity and determinism, and compare across models.

Lab:

Create prompt templates for delivery ETA estimation and customer messaging. Tune temperature/top-p and compare outputs across GPT, Claude, and Gemini.

---

Day 3: Embeddings, Semantic Search, and RAG

Topics Covered:

• Understanding Embeddings
  Learn how embeddings represent semantic meaning and enable similarity-based search.

• Semantic Search and Vector Databases
  Explore Azure AI Search and Cosmos DB for semantic and hybrid search capabilities.

• RAG Architecture
  Understand how RAG combines retrieval with generation to provide context-aware responses.

• Hands-on RAG Build
  Build a simple RAG system using LangChain/LangGraph (JavaScript) and Azure OpenAI.

Lab:

Build a RAG system that retrieves delivery logs and traffic updates, then generates contextual responses to customer queries.

---

Day 4: Agentic AI and Autonomous Systems

Topics Covered:

• Agentic AI Concepts
  Understand autonomous agents, their goals, and how they operate in dynamic environments.

• Agentic Patterns and Architectures
  Learn about function calling, memory management, and human-in-the-loop systems.

• Single-Agent Architecture and Scaling
  Explore design principles, limitations, and strategies for scaling agentic systems.

• Hands-on Agentic System
  Build a modular agent using Azure AI Foundry or Semantic Kernel with LangChain and LangSmith.

Lab:

Create an agent that monitors delivery status, reprioritizes orders, and communicates updates. Use LangSmith to trace agent decisions.

---

Day 5: Advanced Architectures and Model Context Protocol (MCP)

Topics Covered:

• Model Context Protocol (MCP) Overview
  Learn MCP's architecture and how it enables context sharing across systems.

• Connecting LLMs to Tools and APIs
  Understand how LLMs interact with logistics APIs, GPS data, and customer service platforms.

• Final Hands-on Project
  Combine RAG, Agentic AI, and MCP to build a fully integrated delivery assistant.

Lab:

Build a mini assistant that uses RAG for data retrieval, agentic logic for decision-making, and MCP for tool integration. Test with simulated delivery scenarios.

---

Repository Structure

GenAI.Foundations/
├── Module01/          # Day 1 content
│   ├── JavaScript/    # JavaScript exercises and solutions
│   └── Python/        # Python exercises and solutions
├── Module02/          # Day 2 content
│   ├── JavaScript/    # JavaScript exercises and solutions
│   └── Python/        # Python exercises and solutions
├── Module03/          # Day 3 content (coming soon)
├── Module04/          # Day 4 content (coming soon)
└── Module05/          # Day 5 content (coming soon)

Each day's folder contains detailed instructions for labs and hands-on exercises.

---

Prerequisites

• Basic programming knowledge (JavaScript or Python)
• Node.js 18+ (for JavaScript exercises)
• Python 3.8+ (for Python exercises)
• Azure OpenAI access
• VS Code with relevant extensions

Getting Started

1. Clone this repository
2. Navigate to the module you want to work on
3. Follow the Instructions.md file in each module's folder
4. Set up your environment variables (see module-specific instructions)

Technologies Used

• Azure OpenAI - LLM provider
• LangChain - AI orchestration framework
• LangGraph - Agentic workflow management
• LangSmith - Observability and tracing
• Azure AI Search - Vector database and semantic search
• Semantic Kernel - AI application development

---

License

This project is for educational purposes.

Contributing

Contributions and improvements are welcome! Please submit a pull request or open an issue.