Meaning
Retrieval-Augmented Generation; an AI architecture that enhances large language models by fetching relevant external knowledge before generating answers.
Three examples from work
Product design
POHow can we prevent the AI customer assistant from making up return policies?
AI EngineerWe will implement RAG using our verified policy documents as the vector source.
Architecture evaluation
LeadIs fine-tuning necessary, or will a RAG pipeline suffice?
ArchitectRAG is much cheaper to maintain and updates in real-time as docs change.
Performance tuning
DevThe retrieval step in our RAG workflow is introducing two seconds of latency.
LeadWe can optimize semantic chunking and add hybrid keyword indexing.
Pronounce as a single word 'rag' or spell it out. RAG prevents hallucinations by grounding LLM responses in real-time company knowledge.