🎬 MISO Master Class — RAG en entornos reales y los desafíos con LLMs

Video Details

Channel: Universidad de los Andes URL: https://www.youtube.com/watch?v=CwWEOIc2hmg Relevance: ⭐⭐⭐⭐⭐


Summary

Academic master class from Universidad de los Andes covering real-world RAG deployment challenges: hallucination in retrieval (retrieving irrelevant passages), embedding model drift, context window management, evaluation methodology (RAGAS framework), and latency SLA compliance. Based on production deployments at multiple enterprises.


PUMA Relevance

Critical for PUMA’s RAG evaluation design. The RAGAS evaluation framework (faithfulness, answer relevance, context precision) provides metrics for PUMA Stage 4 beyond F1-macro. The context window management section informs PUMA’s prompt design: how many retrieved issues to include without exceeding Llama 3.2’s 8K context.


MOCs