NVIDIA Kumo Relational Outperforms Supervised Machine Learning on Relational Data, No Training Required

NVIDIA's Kumo Relational is a pre-trained foundation model that predicts outcomes directly from enterprise relational databases without task-specific training or feature engineering. It outperforms supervised machine learning models on 41 tasks across 15 databases, including beating AutoGluon ensembles by 8% on the SAP SALT benchmark. The model scales to databases with over 500 billion rows and is accessible via NVIDIA's open-source library or NVIDIA NIM.