Arthur Intelligence

Arthur Intelligence, operating as Arthur AI, is a New York-based firm specializing in the performance management of artificial intelligence systems. The company was founded in 2018 by Adam Wenchel, John Dickerson, Liz O’Sullivan, and Priscilla Alexander with the mission to make AI more transparent, equitable, and observable. The founders' collective experience identified a critical gap in the AI infrastructure for monitoring and ensuring the reliability of production models. Wenchel, the CEO, previously led the AI division at Capital One after it acquired his cybersecurity startup, Anax Security. This experience at enterprise scale, managing high-impact models for credit and fraud, directly highlighted the need for robust AI monitoring solutions. Chief Scientist John Dickerson is a tenured professor of Computer Science at the University of Maryland with a Ph.D. from Carnegie Mellon, whose research intersects machine learning and economics. Liz O’Sullivan, a responsible AI advocate, brought industry experience from leading commercial adoption of AI and now serves as CEO of Vera AI. Priscilla Alexander also came from Capital One, where she led the development of key AI projects.

The company has secured significant financial backing, raising a total of $60.3 million through several funding rounds. This includes a $3.3 million Seed round in December 2019, a $15 million Series A in December 2020, and a $42 million Series B in June 2022. The Series B round, co-led by Acrew Capital and Greycroft, gave the company a post-money valuation of $154.35 million. These funds are aimed at scaling operations to meet surging demand and advancing research and development.

Arthur AI provides a platform that monitors, measures, and improves the performance of machine learning models across their entire lifecycle. The platform caters to enterprise clients, including those in financial services, healthcare, and government, such as Humana and the U.S. Department of Defense. The business model is centered on providing this AI performance management solution, which helps data scientists, ML engineers, and business leaders ensure their AI systems are accurate, explainable, and fair. The service is delivered through flexible deployment options, including SaaS and on-premises solutions, and is available on AWS and GCP marketplaces.

The core of Arthur's offering is a comprehensive AI control and monitoring system that provides visibility into model performance for tabular data, computer vision (CV), natural language processing (NLP), and large language models (LLMs). A key product, Arthur Shield, acts as a firewall for LLMs to protect against threats like hallucinations, toxic language, data leakage, and prompt injection. The platform is model-agnostic and platform-agnostic, integrating with existing tools to provide real-time alerts, bias detection, and customizable performance monitoring. This helps organizations accelerate model operations, mitigate risks, and establish AI governance frameworks, ultimately saving on operating expenses and increasing model-driven revenues. Recently, the company open-sourced Arthur Engine, a real-time AI evaluation tool that allows developers to run evaluations locally to preserve data security.

Keywords: AI performance management, machine learning observability, MLOps, AI monitoring, model explainability, AI fairness, bias detection, LLM security, AI governance, enterprise AI, natural language processing, computer vision, responsible AI, AI infrastructure, data drift detection, Adam Wenchel, John Dickerson, AI model validation, AI risk management, AI control plane