Large Language Models have revolutionized NLP development with prompt-based approaches, but transitioning from prototypes to production systems presents challenges in modularity, transparency, and data privacy. The presentation covers practical strategies for building future-proof NLP pipelines in banking and finance, featuring real-world case studies including S&P Global's commodities trading insights system. Key topics include human-in-the-loop distillation techniques, document understanding pipelines, and methods for creating smaller, faster, privacy-compliant models from large generative systems.
29 Impressions