Fireside Chat with Tess Munsie, Chief of Staff at Fundamental | RAISE Summit 2026
Core Problem Identification: While Large Language Models (LLMs) have transformed the handling of unstructured data (text, imagery, code), they struggle with enterprise structured data (rows, columns, transactions, pricing, inventory), creating a "blind spot" for the data that drives the majority of business outcomes.
- LLMs process information sequentially as tokens, whereas tabular data is inherently relational and non-sequential, causing LLMs to fail when column order changes or when deterministic precision is required.
- Enterprises often face "prediction poverty" despite holding decades of structured data, as insights remain trapped due to messy formats and the high manual lift required for pre-processing.
Fundamental's Solution: Nexus (Large Tabular Model): Fundamental has developed Nexus, a model specifically architected for tabular data to unlock value in structured datasets.
- The model is positioned as complementary to LLMs, functioning like the "right side of the brain" for statistics, probabilities, and predictions, whereas LLMs handle the "left side" of language and reasoning.
- Nexus allows for forward-looking decisions across industries, including demand forecasting (energy), predictive maintenance (shipping), and ad inventory pricing (media).
- Unlike LLMs that often require moving data, Nexus operates on a "bring the model to the data" philosophy to accommodate decentralized, messy, and incomplete enterprise datasets.
Security and Deployment Strategy: The company prioritizes enterprise security by deploying directly into customer Virtual Private Clouds (VPCs) via "confidential compute."
- This architecture ensures model weights are protected and, crucially, that customer data never leaves the VPC or is used to train Fundamental's public models.
- The deployment model addresses data sovereignty concerns, particularly for European clients, by allowing local optimization without transferring sensitive trade secrets.
Market Traction and Valuation: Fundamental has exited stealth status with a valuation of $1.45 billion, targeting a broad range of sectors including energy, gas, telecom, shipping, and consumer packaged goods (CPG).
- The company reports that interest is driven by tangible ROI and the ability to reduce the bespoke, brittle nature of traditional machine learning pipelines.
- Nexus replaces the need for task-specific models, offering a single generalized model for multiple use cases to save data scientists time on feature engineering and maintenance.
Performance and Capabilities: The Nexus model claims to reduce the time data scientists spend on complex forecasting tasks from months to days or hours.
- The model leverages pre-training and world knowledge to pick up incremental performance improvements without the high compute requirements of LLMs.
- Use cases range from optimizing known processes to discovering new applications within specific departments after initial proof-of-concept engagement.
Forward-Looking Statements and Applications:
- Fundamental views the Large Tabular Model (LTM) space as following a similar growth trajectory as the LLM market but currently in a nascent phase.
- The company is actively using Nexus to predict sports outcomes as a proof of capability, having correctly predicted Argentina's 2022 World Cup win and currently predicting France to win the current tournament with a 20.7% probability for overall victory.
- Prediction results are published publicly on
soccer.fundamental.techandfootball.fundamentals.techfor transparency.