Conquer RAG Retrieval Errors: 5-Step Tuning Process in Snorkel Flow
Stop struggling with retrieval errors in your RAG system!
In this video, Snorkel AI product lead Marty Moesta shows a step-by-step approach for optimizing retrieval using Snorkel Flow. Boost your RAG performance and get the right answers every time.
Here's what you'll learn:
The two main culprits behind retrieval errors in RAG systems.
A step-by-step guide to tuning your RAG system in Snorkel Flow for maximum accuracy.
How to identify and address targeted error areas to get the right context for your LLM.
Proven techniques for improving document chunking and embedding models for better retrieval.
How Snorkel Flow empowers collaboration between subject matter experts and data scientists to optimize your RAG system.
See more Snorkel Flow product demos here: • Snorkel Flow Demos: See How it Works!
Timestamps:
00:00 Introduction
01:09 Five-Step Process Overview
01:55 Demo Introduction: Shield Healthcare
02:27 Initial Questions and Observations
03:27 Onboarding Data into Snorkel
04:11 Evaluating LLM Quality
05:20 Partnering Domain Experts and Data Scientists
06:10 Techniques for Improving RAG Pipeline
06:39 Document Chunking
07:47 Document Enrichment via Metadata Models
10:34 Techniques for Fine-Tuning Embedding Models
12:21 Gathering Domain Expert Feedback
13:30 Results of Techniques Applied
14:29 Conclusion and Insights
#enterpriseai #rag #snorkelai
