Fundamental's NEXUS model on SageMaker delivers deterministic predictions for tabular data
Pre-trained on billions of tabular prediction tasks, ready to deploy in days.
Deep Dive
Fundamental's NEXUS, a Large Tabular Model built for structured data, is now available on Amazon SageMaker JumpStart. It produces deterministic, reproducible predictions without manual feature engineering, pre-trained on billions of real-world prediction tasks. NEXUS handles up to a billion rows, cross-schema reasoning, and missing data autonomously. Deploy on an ml.p5en.48xlarge (8× NVIDIA H200 GPUs) and use a scikit‑learn compatible API to train and predict in days instead of months.
Key Points
- NEXUS is a deterministic Large Tabular Model pre-trained on billions of structured prediction tasks, available on SageMaker JumpStart.
- Deploys on ml.p5en.48xlarge instances with 8× NVIDIA H200 GPUs, using a scikit-learn compatible API for training and inference.
- Handles up to a billion rows, autonomously cleans missing data, and reasons across multiple tables without manual feature engineering.
Why It Matters
Enterprises can now generate accurate, deterministic predictions from structured data in days, not months, without sacrificing consistency.