Graphon

Graphon AI is a San Francisco, California-based enterprise AI infrastructure company that develops a "pre-model intelligence" software layer intended to sit between raw enterprise data and foundation models. Its platform ingests documents, video, audio, images, logs, machine-generated data and structured databases and converts them into persistent relational memory, so that large language models and autonomous agents can reason over connected information rather than isolated text snippets. The company positions this as an alternative to retrieval-augmented generation (RAG), which it argues surfaces relevant documents but loses the structural relationships between datasets, and to the token limits of conventional context windows.

The technology is based on graphon functions, a mathematical framework for modelling relationships in large, interconnected systems. The company says its layer automatically discovers and maps structural connections across multimodal data streams before a model queries them, creating persistent context that it describes as effectively unlimited and operable at trillion-token scale, independent of any particular model. Stated application areas include enterprise content management, industrial operations analysis, AI agent and copilot workflows, and multimodal reasoning across connected devices such as cameras, smartphones, wearables and industrial systems.

The company emerged from stealth in May 2026 alongside an $8.3 million seed round. It is led by founder and CEO Arbaaz Khan, with co-founders Deepak Mishra (COO) and Clark Zhang (CTO); its founding and advisory teams include former researchers and engineers from Amazon, Meta, Google, Apple, NVIDIA, Samsung AI Center, NASA, MIT and Rivian. Technical advisors include UC Berkeley professor Christian Borgs, who coined the term "graphon," and Jennifer Chayes, dean of UC Berkeley's College of Computing, Data Science, and Society.

Founding story

The company was founded by former big-technology researchers and engineers, with CEO Arbaaz Khan having previously built customer service models at Amazon. It is co-founded by Deepak Mishra, who serves as COO, and Clark Zhang, who serves as CTO. The technical approach draws on graphon theory developed by UC Berkeley researchers Christian Borgs and Jennifer Chayes, who advise the company.

Business model

Graphon AI sells an infrastructure/middleware layer to enterprises rather than a foundation model, operating between customers' raw data and third-party models. It works with large enterprise customers directly, as illustrated by production deployments with the South Korean conglomerate GS Group, and states plans to expand toward Fortune 500 enterprises, agentic workflows and on-device AI.

Traction

The company reports early production deployments with South Korean conglomerate GS Group, covering customer movement mapping in convenience stores and retail environments and construction site safety monitoring via CCTV analytics. It also cites participation in an SE Ventures accelerator and hackathon wins.

Latest developments

In May 2026 the company emerged from stealth with $8.3 million in seed funding led by Arvind Gupta of Novera Ventures, with participation from the Perplexity Fund, Samsung Next, GS Futures, Hitachi Ventures, Gaia Ventures, B37 Ventures and Aurum Partners (the investment vehicle affiliated with the San Francisco 49ers ownership group). The proceeds are earmarked for building out the relational representation platform, advancing the continuous multimodal graph technology, scaling infrastructure for enterprise AI agent frameworks, and hiring across engineering, machine learning, marketing and go-to-market roles.