Whenever you're ready, walk me through how you'd approach this.
How do you approach fine-tuning versus RAG (Retrieval-Augmented Generation) when building domain-specific agents for engineering data?
Explain how you would choose between the approaches, identify the data and evaluation requirements, and describe when a hybrid architecture is appropriate. Include a practical implementation outline that addresses retrieval quality, hallucination risk, freshness, latency, privacy, and operational maintenance.