Agentic AI for Developers

Goal: Start building structured LLM applications using Lang Chain and Lang Graph.

Lang Chain Basics

• Lang Chain philosophy and architecture
• Core components: LLM, Prompt Template, Chains, Memory
• Using different LLMs interchangeably (LLaMA, ChatGPT, Gemini, etc.)

Lang Graph Introduction

• Lang Graph versus traditional Lang Chain chains
• Thinking in graphs: branching, retries, dynamic paths
• Building single-turn chains and multi-step conversation flows
• Producing structured outputs
• Rapid UI prototyping with Streamlit or Gradio
• Debugging and monitoring with Lang Smith

Hands-on Exercises

• Building a basic Lang Chain agent
• Creating a simple Lang Graph workflow

Goal: Connect LLMs to external knowledge bases effectively.

What is RAG and Why It Is Critical

• Context window limitations and the need for augmentation
• Introduction to vector databases (FAISS, Chroma, Pinecone)
• Metadata filtering strategies

Building a RAG Pipeline

• Retrieval from unstructured data (websites, PDFs, documents, images)
• Chunking strategies and generating embeddings
• Indexing into vector stores
• Retrieval, re-ranking, and query rewriting

Hands-on Exercises

• Building a RAG system over custom PDFs or text files
• Testing retrieval quality and evaluating chunking strategies

Goal: Connect LLMs with real-world APIs, tools, and databases.

Lang Chain Tools

• What are tools? Wrapping APIs and Python functions
• Difference between toolchains and agents

Working with Databases

• Connecting to SQL and NoSQL databases
• Creating dynamic database query tools using LLMs

Multi-Tool Agents

• Dynamic tool selection
• Planning and decision-making with multiple tools

Hands-on Exercises

• Building a simple API tool (e.g., weather connector)
• Building a database query tool (e.g., for employee records)
• Building an agent that dynamically selects and uses multiple tools

Goal: Bring all concepts together into a complete application.

Final Hands-on Project

• Example project ideas:
  • Company knowledge chatbot (retrieval plus tool calling)
  • Travel assistant bot (weather and hotel APIs)
  • HR assistant (RAG over employee data with SQL queries)

Best Practices for Real-World LLM Applications

• Prompt engineering techniques for RAG systems
• Memory management strategies
• Error handling and fallback mechanisms
• Streaming responses from LLMs
• Using function-calling models (Open AI Tools)
• Observability and monitoring with Lang Smith

Goal: Understand MCP as a standard for tool and data integration across LLMs.

What is MCP? (Motivation and history — why interoperability is needed)

How MCP standardizes communication between LLMs and external tools

MCP vs. LangChain Tools/Agents — differences and complementarity

Current adoption: OpenAI, Anthropic, LangChain ecosystem, IDEs

Practical examples:

• Defining a tool with MCP
• Using MCP with databases, APIs, or custom functions

Why MCP matters for enterprise adoption (portability, vendor-agnostic pipelines