Overview

This course introduces the powerful concept of Retrieval-Augmented Generation (RAG), a technique used to optimize the performance, accuracy, and cost of generative AI systems. Focused on building AI pipelines with LlamaIndex, Deep Lake, and Pinecone, this course will equip you with the skills to create robust AI models capable of handling complex datasets and delivering traceable, context-aware outputs. You will explore how to scale RAG pipelines, implement strategies to minimize hallucinations, and improve response accuracy across multimodal AI systems. By the end of the course, you will have hands-on experience optimizing these systems for real-world applications, empowering you to enhance decision-making and operational efficiency. What sets this course apart is its unique combination of theory and practical implementation. By working with cutting-edge tools like LlamaIndex and Pinecone, you'll understand how to balance cost, performance, and accuracy, while gaining insight into the broader context of AI pipelines and decision-making. This course is ideal for data scientists, AI engineers, and MLOps professionals who are looking to expand their expertise in RAG and generative AI. A basic understanding of machine learning concepts is recommended, as the course builds on these foundations to explore more advanced techniques.

Syllabus

Why Retrieval Augmented Generation?

In this section, we explore Retrieval Augmented Generation (RAG) frameworks, focusing on naive, advanced, and modular configurations. We implement Python-based RAG systems for improved AI accuracy and adaptability.

RAG Embedding Vector Stores with Deep Lake and OpenAI

In this section, we cover building and managing RAG pipelines with Deep Lake and OpenAI for efficient AI data handling.

Building Index-Based RAG with LlamaIndex, Deep Lake, and OpenAI

In this section, we explore index-based RAG pipelines using LlamaIndex, Deep Lake, and OpenAI to enhance traceability, precision, and control in AI-driven data retrieval and generation.

Multimodal Modular RAG for Drone Technology

In this section, we explore multimodal modular RAG for drone technology, integrating text and image data retrieval, generation, and performance evaluation using LLMs and MMLLMs.

Boosting RAG Performance with Expert Human Feedback

In this section, we explore adaptive RAG with human feedback loops, focusing on improving retrieval quality and integrating expert input.

Scaling RAG Bank Customer Data with Pinecone

In this section, we explore scalable RAG techniques for bank customer data using Pinecone and OpenAI. Key concepts include EDA, vector scaling, and AI-driven recommendations to reduce churn.

Building Scalable Knowledge-Graph-Based RAG with Wikipedia API and LlamaIndex

In this section, we explore building scalable RAG systems using knowledge graphs, implementing the Wikipedia API, populating a Deep Lake vector store, and constructing a LlamaIndex knowledge graph for semantic search.

Dynamic RAG with Chroma and Hugging Face Llama

In this section, we explore dynamic RAG using Chroma and Llama, focusing on embedding and querying temporary data for real-time decision-making with open-source tools.

Empowering AI Models Fine-Tuning RAG Data and Human Feedback

In this section, we explore RAG data reduction through fine-tuning, focusing on preparing JSONL datasets and evaluating model performance with OpenAI metrics for improved accuracy and cost-effectiveness.

RAG for Video Stock Production with Pinecone and OpenAI

In this section, we explore RAG pipeline implementation for video generation, embedding video comments in Pinecone, and enhancing labels with GPT-4o analysis for efficient video stock production.