Tom Kaltofen: Rethinking feature engineering with mloda

Rethinking feature engineering with mloda Tom Kaltofen mloda is a new open-source Python framework that revolutionizes feature engineering by separating processes (transformations) from data. In traditional workflows, data scientists often rebuild the same feature logic for each environment (development, training, production) and each tool, leading to duplication and inconsistency. mloda tackles this pain by enabling a “write once, reuse everywhere” approach. This decoupling of feature logic from data not only ensures consistency and scalability, but also allows sharing transformation patterns without exposing sensitive business data. This enables fostering a community of reusable feature plugins. In this talk, we’ll explain what mloda is, the problems it solves, and how its approach differs from traditional feature engineering. A live demo will showcase how one mloda plugin can replace repetitive pipeline code, transforming how features are defined, generated, and used in your ML workflows. Talk room 2 (200 people) Sun 10:05 am - 10:35 am