Agentic AI

Module 1: Python for AI & LLM Fundamentals (6 Hours)

Topics:

• Python Recap: Functions, Classes, Decorators, Async/Await

• Introduction to Large Language Models (LLMs): GPT-4, Claude, Gemini, Llama

• Transformer Architecture & Attention Mechanism (Conceptual Overview)

• OpenAI API: Chat Completions, Embeddings, and Function Calling

• Prompt Engineering: Zero-shot, Few-shot, Chain-of-Thought, System Prompts

• Tokens, Temperature, Top-p, and Model Parameters

Hands-on: Build a simple Q&A chatbot using the OpenAI API with Python.

Outcome: Students understand LLM fundamentals, use OpenAI API, and apply prompt engineering techniques.

Module 2: LangChain Framework – Chains, Memory & RAG (9 Hours)

Topics:

• LangChain Architecture: LLMs, ChatModels, Prompts, Output Parsers

• LangChain Expression Language (LCEL) and Runnables

• Memory Modules: ConversationBufferMemory, SummaryMemory, VectorStoreMemory

• Document Loaders: PDF, Web, CSV, Notion, GitHub

• Text Splitters, Embeddings, and Vector Stores (FAISS, Chroma, Pinecone)

• Retrieval-Augmented Generation (RAG): Naive RAG, advanced RAG patterns

• LangChain Tools: Search, Calculator, Python REPL, Custom Tools

Hands-on: Build a document Q&A chatbot with memory using LangChain RAG pipeline over a custom PDF knowledge base.

Outcome: Students can build conversational AI apps with memory and retrieval using LangChain.

Module 3: AI Agents & LangGraph (9 Hours)

Topics:

• What are AI agents? Agent Loop: Observe, Think, Act, Reflect

• ReAct (Reasoning + Acting) Agent Pattern with Tool Integration

• OpenAI Tool Calling / Function Calling Agents

• Introduction to LangGraph: Nodes, Edges, State, and Graph Compilation

• LangGraph StateGraph: Typed State, Reducers, Conditional Edges, Loops

• Agent Memory in LangGraph: Short-term, Long-term, Episodic Memory

• Human-in-the-Loop: Interrupt, Resume, and Approval Workflows in LangGraph

• Debugging & Tracing Agents with LangSmith

Hands-on: Build a ReAct agent with web search, calculator, and code execution tools. Build a stateful customer support agent using LangGraph.

Outcome: Students can build and debug autonomous AI agents and stateful graph-based agent workflows using LangGraph.

Module 4: Multi-Agent Systems & Capstone Project (6 Hours)

Topics:

• Multi-Agent Architecture: Supervisor, Worker, Specialist, and Orchestrator Patterns

• Building Multi-Agent Systems with LangGraph: Subgraphs, Inter-Agent Communication

• CrewAI Framework: Roles, Tasks, Tools, and Agent Crews

• Agent Coordination: Sequential, Parallel, and Hierarchical Execution

• Real-World Multi-Agent Use Cases: Research Automation, Code Review, Content Pipeline

• Deployment: FastAPI + LangServe, Docker, Streamlit UI for AI Agent

• Capstone Project Options:

• AI Research Assistant: Multi-agent system that searches, reads, and summarizes research papers

• Automated Code Reviewer: Agent that reads GitHub PRs and gives structured feedback

• Customer Support Bot: LangGraph-powered agent with RAG, escalation, and ticketing tool integration