Fundamental's NEXUS, a large tabular model, is now available on Amazon SageMaker JumpStart, enabling faster and more accurate predictions from structured data. The model is designed to generate deterministic predictions from enterprise datasets in days rather than months. Users can deploy NEXUS through Amazon SageMaker JumpStart, which provides a managed environment for deploying and running predictions. The model is built for structured data analysis and offers key innovations such as deterministic architecture, native tabular understanding, and non-sequential reasoning.
NEXUS is a foundation model developed by Fundamental and is built for tabular data prediction. Unlike large language models, which are designed for text, NEXUS is trained on billions of real-world prediction tasks across structured datasets. This allows it to process numbers, categories, dates, and unstructured text without manual feature engineering. The model is capable of analyzing multi-dimensional relationships in enterprise tables, such as understanding how multiple factors impact customer churn.
Traditional machine learning approaches require extensive feature engineering and model training, often taking 3–6 months for a single use case. In contrast, NEXUS handles data cleanup and feature engineering automatically, with no manual pipeline required. The model is deployed on a dedicated, single-tenant, network-isolated GPU instance within the SageMaker AI managed environment. Users can train and generate predictions using the Fundamental Python SDK, which provides a scikit-learn compatible API.
Source: awsml
