RAG (Retrieval-Augmented Generation)
An AI architecture that retrieves relevant documents from a knowledge base before generating responses, grounding every output in verifiable source material. In medical device compliance, RAG ensures AI-generated claims are traceable to actual regulatory filings and product documentation.
RAG (Retrieval-Augmented Generation) is an AI technique that separates knowledge retrieval from text generation. Instead of relying on what the model "knows" from training data, RAG first searches a curated document corpus for relevant information, then generates responses grounded in those specific documents.
Why RAG matters for compliance
General-purpose LLMs hallucinate — they generate plausible-sounding but fabricated regulatory facts. RAG eliminates this by constraining the AI to only cite information that exists in your verified document corpus. Every compliance claim traces back to a specific datasheet, certificate, or filing.
RAG vs. fine-tuning
Fine-tuning trains a model on your data, embedding it into the model's weights. RAG retrieves your data at query time. For compliance, RAG is preferred because the source documents change (certificates expire, standards update) and the audit trail requires document-level citations.