Chip Huyen · The LLM sandwich -- the data layer before and after LLMs

ai.bythebay.io Nov 2025, Oakland, full-stack AI conference As LLMs are incorporated into production pipelines, we look into what happens before and after LLMs in a production stack. This talk discusses how to extract necessary data and construct relevant contexts in real time for LLMs to generate the correct responses, and how to monitor those responses to minimize risks and improve user experience.