RAG++ : From POC to Production
Practical RAG techniques for engineers: learn production-ready solutions from industry experts to optimize performance, cut costs, and enhance the accuracy and relevance of your applications.
Practical, tested solutions to getting better higher accuracy out of your POC apps.
Systematic RAG evaluation techniques.
Best practices for consistent and reliable outputs while minimizing hallucination.
Cohere credits to run course notebooks.
Welcome to the course
Weave and Cohere credits set up
Chapter goals
Notebook 1: Baseline RAG Pipeline
From basic to advanced RAG
Wandbot
20/80 rule
RAG best practices
Challenges and solutions
Chapter goals
Evaluation basics
Notebook 2: evaluation
Evaluating retrievers
LLM as a judge
Assertions
Traditional NLP limitations
LLM evaluation in action
Re-evaluating models
LLM eval limitations
Pairwise evaluation
Conclusion
Chapter goals
Data ingestion
Data parsing
Chunking
Metadata management
Data ingestion challenges
Best practices
Notebook 3
Chunking in practice
BM25
Conclusion
Chapter goals
Key techniques
Enhancing context
LLM in query enhancement
Query enhancement case study: Wandbot
Notebook 4: Query enhancement
Chapter goals
Limitations
Compare evaluations
Query translation
Retrieve with CoT
Metadata filtering
Logical routing
Context stuffing
Cross encoder
Notebook 5: retrieval and reranking
Reciprocal rank fusion
Hybrid retriever
Weaviate Vector Database
Weaviate Hybrid Search
Conclusion
This course is for people with:
familiarity with Python
basic understanding of RAG