LLM with Python
DAY 1 – Building a Retrieval-Augmented Generation
(RAG) System
Goal: Understand context retrieval, embeddings, and create a working RAG chatbot using
open-source LLMs.
Time Topic Learning Outcomes Hands-On Activities
9:00 – LLM Ecosystem - Understand the Quick demo: Using
9:30 AM Overview architecture of modern LLM ollama to run a local
apps model
- Difference between model,
retriever, and orchestrator
- Where APIs (OpenAI,
Ollama, etc.) fit in
9:30 – Understanding - Convert text into Hands-on: Build a
10:30 AM Embeddings and embeddings simple document search
Vector Databases - Store and retrieve data engine
from FAISS / Chroma /
Milvus
10:30 – ☕ Break
10:45 AM
10:45 – Designing the RAG - Learn how RAG improves Draw architecture and
12:15 PM Architecture factual accuracy run retrieval pipelines
- Understand chunking,
retrieval, context window,
and re-ranking
12:15 – 🍴 Lunch Break
1:00 PM
1:00 – Building a RAG - Use LangChain / Project: “Chat with
3:00 PM Chatbot in Python LlamaIndex with local Company Policy” bot
LLMs (e.g., Mistral, Gemma) using your own
- Query over PDF or text documents
corpus
- Manage memory and
streaming responses
3:00 – ☕ Break
3:15 PM
3:15 – Enhancing the RAG - Add context ranking, Deploy chatbot UI +
5:00 PM Pipeline caching, and evaluation measure retrieval quality
- Integrate Streamlit/Gradio
for UI
Day 1 Outcomes
● Understand how to build a production-style RAG pipeline
● Learn to use LangChain, FAISS, ChromaDB, sentence-transformers
● Deliverable: Working “Chat with Documents” app using local or API LLM
DAY 2 – Building Agentic and MCP-driven AI Systems
Goal: Create autonomous LLM agents, connect them with tools and APIs, and explore
Model Context Protocol (MCP) for composability.
Time Topic Learning Outcomes Hands-On Activities
9:00 – From Chatbots to - Understand agents, tools, Demo: Agent reasoning
9:45 AM Agentic AI reasoning loops chain visualization
- Learn how OpenAI
function calling and ReAct
patterns work
9:45 – Building a Tool-Using - Connect your LLM with Code Lab: “Smart
11:15 AM LLM Agent external tools (search, Research Assistant” agent
calculator, APIs) (uses DuckDuckGo +
- Use LangChain or Wikipedia)
LlamaIndex agents
11:15 – ☕ Break
11:30 AM
11:30 – Introduction to MCP - Understand what MCP is Explore MCP architecture
12:45 PM (Model Context (as used by Anthropic, with examples
Protocol) LangChain, etc.)
- Learn how to make LLMs
interact modularly across
tools
- Explore structured context
exchange
12:45 – 🍴 Lunch Break
1:30 PM
1:30 – Hands-on: Building - Build a pipeline Project: “Research Copilot”
3:15 PM MCP-Style Modular connecting LLM + Retrieval — RAG + Tool + Planner
LLM System + API + Memory
- Each component
communicates via a context
protocol
3:15 – ☕ Break
3:30 PM
3:30 – Deploying & Scaling - Use Streamlit/FastAPI for Deploy one of your agents
5:00 PM LLM Apps UI locally (Streamlit UI + API
- Use OpenVINO or endpoint)
Ollama for local inference
- Design modular
architecture for real-world
LLM apps
Day 2 Outcomes
● Understand Agentic AI architecture (ReAct, Toolformer, function-calling)
● Build a multi-tool LLM agent with reasoning loops
● Learn Model Context Protocol (MCP) for interoperable LLM pipelines
● Deploy working LLM apps via Streamlit or FastAPI