Beyond the lakehouse: Fundamental's NEXUS bypasses manual ETL with a native foundation model for tabular data

2026-02-05 · Source: VentureBeat · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics, Cloud Computing & IT Infrastructure · Depth: Intermediate, medium

Summary

Fundamental, an AI firm co-founded by DeepMind alumni, has launched with $255 million in funding to introduce NEXUS, a Large Tabular Model (LTM). This LTM is designed to process structured, relational business data, such as that found in ERP and CRM systems, which traditional Large Language Models (LLMs) struggle with due to their sequential logic and tokenization methods for numbers. NEXUS, trained on billions of real-world tabular datasets using Amazon SageMaker HyperPod, bypasses manual feature engineering by directly ingesting raw tables and identifying latent non-linear patterns. The model aims to provide predictive intelligence for split-second decisions, such as fraud detection or equipment failure forecasts, and is commercially available via AWS Marketplace, allowing deployment within customer environments with existing AWS credits and full encryption for data privacy.

Key takeaway

For VPs of Data or Engineering evaluating AI solutions for business forecasting, NEXUS offers a direct path to predictive intelligence for tabular data, bypassing labor-intensive ETL and feature engineering. Your teams can deploy this encrypted model within your AWS environment using existing credits, enabling automated, high-speed decision-making for critical applications like fraud detection or supply chain optimization without moving sensitive data.

Key insights

NEXUS is a Large Tabular Model designed to natively understand and predict from complex, non-sequential business data.

Principles

• Tabular data is inherently non-sequential.
• LLMs struggle with numerical tokenization.
• Order-invariance is critical for tabular data.

Method

NEXUS ingests raw tables directly, identifying latent patterns across columns and rows without manual feature engineering, and returns regressions or classifications into the enterprise data stack.

In practice

• Predict credit card fraud in real-time.
• Forecast equipment failures from sensor data.
• Anticipate hospital readmission risks.

Topics

• Large Tabular Models
• Tabular Data Processing
• Foundation Models
• Predictive Analytics
• Enterprise AI

Best for: VP of Engineering/Data, Executive, Investor, Data Scientist, MLOps Engineer, CTO

Related on AIssential

• Fundamental’s Large Tabular Model NEXUS is now available on Amazon SageMaker JumpStart — NEXUS is a foundation model purpose-built for tabular data, delivering deterministic, scalable predictions without extensive feature engineering.
• Startup Introduces 'Large Tabular Model' for Spreadsheet Data — Nexus is a specialized foundation model for tabular data, addressing LLM limitations in structured data analysis.
• ReSS: Learning Reasoning Models for Tabular Data Prediction via Symbolic Scaffold — ReSS combines symbolic decision trees with LLMs to generate accurate, faithful, and explainable tabular data predictions.
• Introducing TabFM: A zero-shot foundation model for tabular data — TabFM applies zero-shot in-context learning to tabular data, eliminating manual tuning and feature engineering for classification and regression.
• SE Radio 724: Jure Leskovec on Relational Graph and Foundational Models — Relational deep learning transforms enterprise tabular data into graphs for superior, automated predictive AI.
• LLMs on Tabular Data with Limited Semantics: Evidence from Industrial Car Retrofit Prediction — LLMs serve best as complementary components to strong tabular machine learning baselines, not replacements, for industrial data with limited semantics.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by VentureBeat.