Agentic AI Course by Andrew Ng

📚 Course Overview

This repository contains my learning materials, assignments, and projects from the Agentic AI course by Andrew Ng and DeepLearning.AI.

⭐ Course Review

This course is an excellent resource for updating your knowledge about Agentic AI with well-structured documentation by Andrew Ng.

My Experience:

• The first module starts a bit slow, providing foundational concepts
• The pace picks up in later modules, covering material more quickly
• Excellent coverage of best practices in Agentic AI development
• Well-organized content with practical, hands-on assignments
• Worth every penny - highly recommended for anyone interested in building AI agents

📖 Course Structure

The course is divided into 5 modules covering key agentic AI design patterns:

Module 1: Planning & Research Agent

• Building a reflective research agent with FastAPI
• Multi-step agent workflow (planner → research → writer → editor)
• Tool integration: Tavily search, arXiv, Wikipedia APIs
• Full-stack web application with Postgres database
• Task orchestration and progress tracking

Module 2: Reflection Design Pattern

• Iterative improvement through LLM-based reflection and critique
• Flashcards Agent: Generate and improve study materials using reflection
• PII Removal: Defensive reflection for protecting sensitive information
• SQL Agent: Refine database queries through self-evaluation
• Visualization Agent: Enhance data charts using iterative critique
• Comparing reasoning vs general-purpose models for reflection

Module 3: Tool Use Design Pattern

• Function calling and tool integration paradigms
• OpenAI function calling vs AISuite approaches
• Email Agent: Focused tool implementation with FastAPI UI
• SQL Agent: Database interaction with dynamic schema exploration
• Research Agent: External API tools for web search and academic research
• Introduction to Model Context Protocol (MCP) for future tool standards

Module 4: Practical Tips for Building Agentic AI

• Evaluations(evals) and Error Analysis for Agentic AI
• Component-level evaluations
• End to end evaluations
• Latency, cost optimization

Module 5: Multi-agent Collaboration

• Dividing complex tasks across multiple specialized agents
• Customer Service Pipeline: Sequential multi-agent orchestration
  • Planner → Coder → Executor → Reflector architecture
  • Error recovery and self-correction capabilities
• Agent communication patterns (linear sequences vs concurrent dialogues)
• Final Project: Complete agentic workflow system
  • Planning agent coordinates research, writing, and editing
  • External tool usage and feedback integration
  • Production-ready research report generation

🗂️ Repository Contents

├── Lecture_notes/          # PDF lecture materials for all modules
├── Module1/                # Introduction and planning agents
├── Module2/                # Tool-calling agents (SQL, visualization)
├── Module3/                # Advanced agent techniques
├── Module4/                # Design patterns
├── Module5/                # Capstone project
└── certificate.png         # Course completion certificate

Each module contains:

• Ungraded Labs: Practice exercises and examples
• Graded Labs: Assessed practical work
• Assignments: Module assessments with solutions
• README files: Additional documentation and instructions

🛠️ Technologies Used

• Python - Primary programming language
• LangChain & AISuite - Agent frameworks and tool orchestration
• OpenAI API - Language models and function calling
• FastAPI - Web application framework
• PostgreSQL - Database for state management
• Jupyter Notebooks - Interactive development and assignments
• Docker - Containerization for deployment
• External APIs - Tavily (web search), arXiv (academic papers), Wikipedia
• Pydantic - Data validation and structured outputs

🎓 Key Learnings

• Reflection Pattern: Building self-improving AI systems through critique and iteration
• Tool Use Pattern: Integrating external tools and APIs with LLM function calling
• Multi-agent Collaboration: Orchestrating specialized agents for complex workflows
• Planning Agents: Breaking down tasks and coordinating execution steps
• Defensive Strategies: Protecting against prompt injection and data leakage
• Production Deployment: Building production-ready agentic systems with FastAPI and Docker
• Best Practices: Error handling, iterative refinement, and modular agent design

🚀 Getting Started

Each module contains its own README with specific setup instructions. Generally:

1. Clone this repository
2. Navigate to the desired module
3. Install dependencies: pip install -r requirements.txt
4. Follow module-specific instructions

📝 Notes

This repository is for educational purposes and contains my personal work and notes from the course. The course materials are property of DeepLearning.AI.

🔗 Course Link

Agentic AI by DeepLearning.AI

---

Instructor: Andrew Ng
Platform: DeepLearning.AI
Completion Date: November 2025