Retrieval Augmented Generation (RAG) – AI Engineering

What is Retrieval Augmented Generation (RAG) ?

Retrieval Augmented Generation (RAG) is an AI technique that improve Large Language Model (LLM) responses by fetching facts from external knowledge bases.

What happens without RAG ?

Sometimes AI give Wrong Answers confidently and those answers look convincing as well. This happens because of AI Hallucination.

How RAG Solves this Problem ?

Instead of relying only on its trained data, the AI searches a custom database, retrieves relevant information, and uses it to generate accurate, upto-date, and verifiable answers. Simply AI doesn’t answer immediately. It first searches for relevant information.

How RAG Works ?

RAG acts as an Open Book Exam.

Without RAG: Answer directly comes from memory.

With RAG: Search Documents, Read relevant Documents and then gives answer. Which is much more reliable.

RAG accomplishes this through following stages :

STEP 1 : Collect Documents

Authorized external data like company documents, PDF’s, or Databases will become the knowledge source

STEP 2 : Chunking

LLMs cannot efficiently process huge documents.

Therefore, Large documents are broken down into small chunks. Instead of storing an entire PDF / Document, We store smaller pieces.

will continue …