
RAG Pipeline: 7 Iterations Explained!
Cyril Imhof
This hands-on tutorial guides you through creating an Agentic Retrieval-Augmented Generation (RAG) pipeline using NVIDIA's open Nemotron models for embedding and reranking. 🔍 What You'll Learn in this RAG Tutorial: 1️⃣ NVIDIA Nemotron RAG Models: Access and use open Nemotron RAG models from Hugging Face for retrieval. 2️⃣ Custom RAG Pipeline: Build a full workflow in LangChain with data loading, chunking, and a Contextual Compression Retriever. 3️⃣ Agent Development: Use Nemotron Nano 2 9B as your LLM and connect retrieval tools into an agent with create_react_agent. 4️⃣ Live Demo: Test the RAG agent on complex queries for accurate, grounded answers. #RAG #AIAgent #Nemotron #NVIDIA 📝 Tech blog: https://nvda.ws/47iK9El 🤗 Model: https://nvda.ws/3JmaSrC Access more NVIDIA Nemotron developer resources and join our developer community: ⬇️ Developer Resources → https://nvda.ws/425fFUJ 📚 Explore Models & Datasets → https://nvda.ws/4n9Ad6N 👥 Join the Community → https://nvda.ws/46Rxucr 💻 Visit the Nemotron Discord channel→https://nvda.ws/421EzEC ▶️ Watch Tutorials & Livestreams → https://nvda.ws/4n5WrXo 🗳️ Share your ideas and vote on Nemotron features → https://nvda.ws/4qbxX0L 0:00 - Intro to Agentic RAG: NVIDIA Nemotron Models for Customizable AI 0:13 - Deep Dive into Nemotron RAG Models on Hugging Face (Embedding & Reranking) 0:39 - Pipeline Setup: Nemotron Nano 2 -9B & Custom Hugging Face Model Classes 1:28 - Data Ingestion, Chunking, and Vector Store (FAISS) 2:11 - Creating the Contextual Compression Retriever 2:29 - Agent Creation: Tool Wrapping and System Prompt 2:53 - Live Agent Demo: Testing Password Requirements Query 3:04 - Next Steps & Resources: Blog, Workshop, and Deployment

Cyril Imhof

Mehul Mathur

Nidhi Chouhan

Nidhi Chouhan

Daksh Rathore

Vikash Kumar

LOUIS PYTHON

Data Science Gems