Graphon AI has emerged from stealth with an $8.3 million seed funding round. The company aims to build a new data layer for enterprise AI that fills a major gap in current models. Their goal is to help large organizations connect and understand their complex, multimodal data more effectively. This layer sits before the foundation models, enabling better reasoning across diverse data sources.

The Math Behind the Innovation

The company is named after a mathematical concept called a graphon, which was co-invented by its advisors. A graphon is a way to capture the structure of relationships in large networks, like social graphs or data connections. It exists at the intersection of pure mathematics and computer science. Graphon AI is turning this concept into a practical software tool.

Foundation models today can process about one million tokens at a time, but companies have trillions of tokens stored in documents, videos, images, and logs. Current solutions can retrieve relevant data but struggle to discover hidden relationships across different data types. Graphon aims to change that by creating a persistent relational memory that captures these connections.

How the Product Works

The system uses graphon functions to analyze multimodal data streams and automatically identify relationships. It produces an organized map of how different data points relate to each other, which can be queried by large language models or other AI agents. This approach allows organizations to reason across their entire data ecosystem without being limited by the model’s context window.

This means enterprises could ask complex questions involving multiple data sources and receive insights that were previously impossible. For example, a company could link surveillance footage, compliance logs, and customer records seamlessly. The goal is to provide a foundational layer that enhances the reasoning power of any AI system working with enterprise data.

An Expert-Backed Approach

Leading academics from UC Berkeley, including Jennifer Chayes and Christian Borgs, serve as advisors. They helped formalize the mathematical idea behind graphons. The company’s founders include Arbaaz Khan, Deepak Mishra, and Clark Zhang, with experience from Amazon, Meta, Google, Apple, NVIDIA, and NASA. Their team combines research and industry expertise to develop this novel solution.

Investors are equally diverse. The seed round was led by Novera Ventures, with participation from firms like Perplexity Fund, Samsung Next, Hitachi Ventures, and others. Notably, South Korean conglomerate GS Group, which is also an early customer, invested. The mix of investors shows that many industries face the same data challenges, from tech to manufacturing to retail.

In summary, Graphon AI is building a new foundation for enterprise data that could revolutionize how organizations connect and reason across their complex data landscapes. With strong academic backing and strategic investors, the company is positioned to address a major gap in AI infrastructure. Its technology could unlock new levels of insight and automation across many sectors.

Inspired by

• https://thenextweb.com/news/graphon-ai-pre-model-intelligence-layer-seed-funding