RAG App Development - up1/training-courses GitHub Wiki

🧠 RAG App Development Course Outline

Target Audience: Developers, AI Engineers, Solution Architects
Duration: 1–2 Days (Adjustable)
Format: Theory + Demos + Hands-on Labs

🗓️ Day 1 – Introduction & Core Concepts

🔹 Module 1: Introduction to RAG

What is Retrieval-Augmented Generation?
Why RAG? Limitations of traditional LLMs
RAG Architecture Overview
Key Components: Query, Retriever, Vector DB, Generator

🔹 Module 2: Core Components Deep Dive

Vector Embeddings: Converting text to vectors
- Providers: OpenAI, Cohere, HuggingFace
Document Chunking & Preprocessing Techniques
Vector Stores: FAISS, Pinecone, ChromaDB
LLM Integrations: OpenAI, Claude, Ollama

🔹 Module 3: Building Blocks of a RAG App

Step-by-step RAG pipeline:
- Data ingestion
- Indexing documents
- Query understanding
- Retrieval
- Generation
Types of RAG:
- Retrieve-then-Generate
- Generate-then-Retrieve

🔹 Lab 1: Build a Basic RAG App

Dataset: PDF/Text/Markdown
Embeddings + FAISS + OpenAI GPT
Simple frontend (Streamlit or Flask)

🗓️ Day 2 – Advanced Features & Productionizing

🔹 Module 4: Enhancing RAG Performance

Prompt engineering for better generation
Hybrid search (semantic + keyword)
Metadata filtering
Response re-ranking
Caching strategies

🔹 Module 5: Evaluation & Metrics

Groundedness & Faithfulness
Evaluation methods: BLEU, ROUGE, LLM-as-a-judge
Human-in-the-loop validation

🔹 Module 6: Scaling & Deployment

Using LangChain, LlamaIndex, Ragas for modular RAG
Integrating APIs (e.g., LangSmith)
Security and privacy considerations

🛠️ Tools & Libraries

Embeddings: OpenAI, Cohere, HuggingFace
Vector DBs: FAISS, ChromaDB, Pinecone
Frameworks: LangChain, LlamaIndex
Deployment: Flask, Streamlit, Docker
Evaluation: RAGAS, LLM-judge tools

✅ Outcomes

By the end of the workshop, participants will:

Understand the RAG architecture
Build and deploy a basic RAG-powered app
Explore advanced optimization techniques
Evaluate and improve RAG-based systems