Story Is All You Need - Transforming Structured Data into Narratives for LLM-Based Predictive Analytics

Explore a revolutionary approach to predictive analytics through this 24-minute conference talk that challenges traditional machine learning methodologies by introducing the "data-as-story" paradigm. Learn how to transform structured data such as transaction logs and app usage data into coherent, text-based narratives that can be fed directly into Large Language Models for generating predictive scores, effectively bypassing conventional modeling and feature engineering processes. Discover through real-world case studies from Wealthsimple how this storytelling technique can predict major life events from credit card transactions and detect fraud anomalies from app behavior narratives, often achieving superior performance compared to rule-based systems. Master a practical framework for implementing the "data-as-story" technique while understanding how storytelling can streamline MLOps pipelines and leverage LLMs for predictive analytics tasks traditionally handled by specialized models. Gain insights into unlocking new predictive capabilities and operational efficiency in real-world applications, presented by Lin Liu, Director of Data Science at Wealthsimple, during the MLOps World GenAI Summit 2025.