Learn how to integrate LLMs into production-grade machine learning pipelines that blend structured and unstructured data in this 21-minute talk by Andrew Moreland, Co-Founder of Chalk. Discover techniques for using real-time feature engineering to surface relevant results in applications like fraud detection, recommender systems, and inference with minimal latency. Explore methods for optimizing speed and efficiency by dynamically adjusting search complexity based on structured data cues. Gain insights on leveraging LLM retrieval at scale and implementing continuous evaluation processes to improve recommendations over time. Andrew brings valuable experience from building large government data infrastructure projects at Palantir and co-founding Haven Money (acquired by Credit Karma).