Companies are increasingly showing interest in AI and its potential to enhance productivity. However, they remain cautious about associated risks. In a recent survey by Workday, enterprises identified timeliness and reliability of underlying data, potential bias, and security and privacy as the primary obstacles to AI implementation.
To address these concerns, Scott Clark, co-founder of the AI training platform SigOpt (acquired by Intel in 2020), founded Distributional. The company aims to develop software that ensures the safety, reliability, and security of AI. Clark envisions Distribution
al as the modern enterprise platform for AI testing and evaluation, providing a proactive approach to identifying and addressing AI risks before they impact customers in production.
Distributional’s core product focuses on detecting and diagnosing potential harm from large language models and other AI models. It aims to semi-automatically determine what, how, and where to test these models. The software offers organizations a comprehensive view of AI risk in a pre-production environment, similar to a sandbox.
Clark acknowledges the challenges faced by teams in ensuring high-quality AI testing on a regular basis. Distributional’s platform includes an extensible testing framework for continuous testing and analysis of stability and robustness. It also features a configurable testing dashboard to visualize and understand test results, along with an intelligent test suite to design, prioritize, and generate the right combination of tests.
While Distributional is still in the co-design phase with enterprise partners and lacks revenue, Clark believes its differentiator lies in its enterprise focus. He emphasizes building software capable of meeting the data privacy, scalability, and complexity requirements of large enterprises, distinguishing it from existing tools that are often individual developer-focused.
Despite being in the early stages without paying customers, Distributional has secured funding through an $11 million seed round led by Andreessen Horowitz’s Martin Casado, with participation from Operator Stack, Point72 Ventures, and angel investors from SV Angel. Clark envisions the platform generating revenue next year once it launches in general availability and design partners convert to paid customers.
Clark expresses optimism in creating a virtuous cycle for customers, where better testing instills confidence in deploying AI applications. As AI deployment grows, its impact scales, leading to more complex challenges that require additional testing to ensure safety, reliability, and security.
