Generative AI and RAG Coursework

Last Updated: May 18, 2026

This repository contains concept-focused notes and hands-on notebooks covering the core ideas of Generative AI, Retrieval-Augmented Generation (RAG), and advanced agentic workflows.

Directory Structure

01-BERT

Foundational NLP models and architecture.

• bert_fundamentals.ipynb: Tokenization, loading BERT, extracting hidden states, identifying NER, custom PyTorch structures.
• bert_flashcards.html: Visual explanation of BERT encoder blocks, MLM, and NSP.

02-Basic-RAG-and-LLMs

Core concepts for Prompting, generation, and basic similarity matching.

• llm_generation_embeddings.ipynb: Exploring embeddings via APIs (OpenAI/OpenRouter) and generating text via local transformers like GPT-2.
• rag_scratch_implementation.ipynb: Complete RAG loop implementation with pure math (cosine similarity, dot products) and basic API ingestion.

03-Vector-Search

(Coming Soon) Vector search algorithms, ChromaDB implementations, and FAISS.

04-Advanced-RAG

(Coming Soon) LangChain, LangGraph orchestrations, retrieval evaluations, hybrid search, and production GenAI pipelines.

05-Miscellaneous

Support scripts and testing directories.

Lectures

Concept notes physically written and typed for learning.

• rag_01_foundations_and_architecture.pdf: Why RAG exists, core architecture, and the retrieval-generation flow.
• rag_03_query_time_retrieval.pdf: Query-time retrieval, prompt augmentation, and grounded generation.

Learning Goal

This repository focuses on building correct mental models first, followed by lightweight hands-on practice, extending natively into production-grade pipelines.