← EXPERIMENTS
AI · 2024
RAG Implementation
QA system that lets users point to any data source and get contextual LLM responses.
What It Does
A question-answering system that lets users point to any data source - codebases, PDFs, documentation - and get contextual LLM responses.
Stack: LangChain · AWS Aurora (vector store) · Llama 3 (base model)
Why
A foundational experiment that preceded the more complex RAG work at Artha Intelligence. The goal was to understand the full retrieval pipeline from chunking → embedding → retrieval before adding complexity.
You can't debug a broken retrieval system if you don't understand what each stage is doing. Building it from scratch with visible seams - not using a pre-packaged RAG library - was the point.