2025-06-17 –, Main stage
Khiops is an open-source, automated Machine Learning (AutoML) library designed for high-efficiency data science workflows. Developed at Orange over 25 years of R&D, Khiops stands out by offering a unique combination of full automation, interpretability, and scalability, making it a compelling alternative to mainstream ML frameworks.
Unlike many AutoML solutions that focus solely on model selection and hyperparameter tuning, Khiops automates the entire pipeline, including feature engineering, variable selection, and model building, while ensuring results are interpretable and robust against overfitting—all without hyperparameter tuning.
Khiops is particularly powerful for multi-table datasets, a common scenario in real-world applications such as fraud detection, risk assessment, and predictive maintenance. It natively processes relational data structures (e.g., customer transactions, logs, IoT data) and extracts the most relevant predictive patterns automatically, without requiring manual data transformations.
By combining efficiency, interpretability, and ease of deployment, Khiops is an ideal solution for organizations looking to scale AI responsibly and effectively.
Luc-Aurélien Gauthier is a data science expert and open-source advocate, currently contributing to the development and promotion of Khiops, an automated, interpretable, and scalable machine learning library. With extensive experience in AI applied inthe industry, he focuses on making advanced ML techniques both efficient and accessible. Passionate about responsible AI and model interpretability, he actively works on bridging the gap between research and real-world applications.