MLflow Evaluation Pipeline Design

Outline the design of an MLflow evaluation pipeline for your Proxy Analyst AI. How will you log LLM prompts, responses, and metrics (e.g., ROUGE, custom validity scores)? Describe the steps to compare AI-generated summaries and recommendations against a 'golden' dataset using MLflow's tracking and logging capabilities.