1. Tutorial Title

Building a Hands-On Retrieval-Augmented Generation (RAG) System with LangChain, OLLAMA, LAMA 3, FAISS andStreamlit,

2. Tutorial Abstract

This tutorial provides a step-by-step, hands-on approach to building a Retrieval-Augmented Generation (RAG) system using popular AI tools such as LangChain, OLLAMA, LAMA 3, FAISS, and Streamlit. Participants will learn how to design and implement an end-to-end RAG system that efficiently retrieves information from a custom knowledge base and generates insightful responses using advanced natural language generation models. The tutorial is geared towards data scientists, machine learning engineers, and AI practitioners interested in developing interactive, intelligent applications that require sophisticated question-answering and document retrieval capabilities. By the end of the session, attendees will have a fully functional RAG application that integrates seamlessly with a user-friendly interface.

3. Tutorial Format/Timeline

☐ 2 hours
☒ 4 hours (recommended for hands-on coding, testing, and Q&A)
☐ 8 hours

4. Target Audience

This tutorial is designed for:
- Machine Learning Engineers and Data Scientists who want to build intelligent, interactive applications using RAG.
- AI Practitioners and Researchers interested in advancing their understanding of large language models (LLMs), vector databases, and retrieval-based generation.

Prerequisites:
- Intermediate knowledge of Python programming.
- Basic understanding of natural language processing (NLP) and machine learning principles.
- Familiarity with large language models, embeddings, and vector-based search concepts.

Expected Participants: 20-40 attendees who have experience working with Python and basic AI/ML concepts.

5. Presenter(s)

Presenter: Partha Deka

Affiliation: Senior Staff Engineer, Intel Corporation

Contact Information: partha.pritamdeka@gmail.com, partha.deka@intel.com

Linkedin: https://www.linkedin.com/in/parthapritamdeka

Google Scholar: https://scholar.google.com/citations?user=qO_1wJgAAAAJ&hl=env

Book Press release: https://www.odbms.org/2024/09/on-xgboost-for-regression-predictive-modeling-and-time-series-analysis-qa-with-partha-deka/

5.1 Short Bio

Partha Deka is a seasoned Data Science Leader with over 15 years of experience driving innovation across the semiconductor supply chain and manufacturing sectors. Currently serving as a Senior Staff Engineer at Intel Corporation, Partha has led high-impact teams in developing cutting-edge AI and machine learning solutions, resulting in significant cost savings and process optimizations. Among his notable achievements is the development of a computer vision system that dramatically enhanced logistics efficiency at Intel, leading his team to be recognized as a finalist for the prestigious CSCMP Innovation Award.

Before his role at Intel, Partha made significant contributions at General Electric (GE), where he demonstrated his expertise in data science and machine learning. During his tenure, he filed multiple patents, including Delivery Status Diagnosis for Industrial Suppliers Using Machine Learning and Auto Throttling of Input Data and Data Execution Using Machine Learning and Artificial Intelligence. These patents have received over 30 citations, underscoring their impact and importance in the field.

A recognized thought leader in the AI community, Partha is a Senior IEEE Member, a published author, and a regular speaker at industry conferences. He is the author of the book XGBoost for Regression Predictive Modeling and Time Series Analysis, which covers foundational knowledge to advanced applications in XGBoost, including time series forecasting, feature engineering, model interpretability, and deployment techniques. His expertise has been acknowledged through his role as a paper reviewer for the prestigious NeurIPS conference, where he contributes to advancing AI and machine learning research. His work continues to shape the field, particularly in applying advanced analytics to enhance semiconductor manufacturing processes.

6. Previous Tutorials Given by the Presenter(s)

The presenter has previously conducted hands-on tutorials on topics including:
- Introduction to Large Language Models and their Applications: Covered the fundamentals of LLMs with practical examples.
- Building NLP Applications with OpenAI's GPT and LangChain: Introduced attendees to LangChain and GPT for developing NLP applications.
- Advanced Techniques in Text Retrieval and Search Systems: Focused on building and deploying vector-based search systems using FAISS.

This proposed tutorial on RAG with LangChain and other advanced tools represents a novel approach, integrating recent advancements in generative AI and retrieval systems.

7. Significance of the Topic

Retrieval-Augmented Generation (RAG) is an emerging and essential technique that combines the strengths of retrieval-based and generation-based models. As large language models like ChatGPT-4 become more popular, the RAG framework is critical for building more accurate, domain-specific, and interactive AI applications. This tutorial’s approach—using LangChain, FAISS, Streamlit, and ChatGPT-4—provides a timely, hands-on application of these methods, meeting the AI research community’s growing interest in efficient retrieval and knowledge-aware generation models. While RAG systems have been discussed in research, this practical tutorial bridges the gap by enabling attendees to build their own RAG system from scratch.

8. Outline and Description of the Tutorial

1. Introduction to RAG Systems and Key Components (20 minutes)
2. Setting Up the Environment (15 minutes)
3. Building the Retrieval System with FAISS (40 minutes)
4. Implementing the Generation System with ChatGPT-4 (30 minutes)
5. Connecting Components with LangChain (40 minutes)
6. Deploying the RAG Chatbot system with Streamlit (20 minutes)
7. Evaluation the RAG system responses based on questions & Troubleshooting (15 minutes)
8. Q&A (20 minutes)

9. Reading List

1. Lewis, P., et al. (2020). Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks. Proceedings of NeurIPS 2020. Available at: https://arxiv.org/abs/2005.11401
2. LangChain Documentation: https://langchain.com/docs/
3. OpenAI API Documentation: https://platform.openai.com/docs/
4. FAISS Documentation: https://github.com/facebookresearch/faiss
5. Streamlit Documentation: https://docs.streamlit.io/