Founded by Scott Clark, Intel's former general manager of AI softwareDistributionalThe company recently closed a $19 million Series AFinancing, led by Two Sigma Ventures.
In an interview with TechCrunch, Clark said Distributional was inspired by testing issues he encountered while applying AI at Intel and his experience working at Yelp as head of software for the ad targeting division. He emphasized:Â "As the value of AI applications continues to increase, so does the operational risk they pose. Our platform helps AI product teams proactively and continuously detect, understand and address AI risks before they occur."
Clark joined Intel through acquisition.In 2020, Intel acquired SigOpt, a model experimentation and management platform co-founded by Clark.In 2022, Clark was named vice president and general manager of Intel's AI and supercomputing software division. During his time at Intel, Clark and his team were often plagued by AI monitoring and observability issues.
Clark noted that the non-deterministic nature of AI makes it possible for the same inputs to produce different outputs. This, coupled with the fact that AI models are dependent on a number of factors (e.g., software infrastructure and training data), makes it exceptionally difficult to find errors in AI systems.
According to a RAND 2024 survey, more than 80% of AI projects end up failing. Generative AI is particularly challenging for organizations, with Gartner research predicting that one-third of deployments will be abandoned by 2026.
To address these issues, Clark created Distributional, which is designed to simplify AI auditing. The platform automatically creates statistical tests of AI models and applications based on developer specifications and organizes test results in a dashboard. Users can collaborate on test "repositories" in the dashboard, categorize failed tests, and recalibrate tests when necessary.
Distributional provides cross-organizational visibility into what, when, and how AI applications are being tested and how those tests change over time.Clark said, "We provide a repeatable process for AI testing that can be applied to similar applications through the use of shareable templates, configurations, filters, and labels."
While AI lab solutions such as Kolena, Prolific, Giskard and Patronus are already on the market, Clark believes Distributional offers a more "white glove" experience. The company handles customer installations, implementations and integrations, and offers AI testing troubleshooting services.
Distributional, which just closed a Series A round of funding, plans to expand its tech team, focusing on UI and AI research engineering.Clark expects the company's headcount to grow to 35 employees by the end of the year, while the company begins its first enterprise deployments.
In addition to Two Sigma Ventures, Andreessen Horowitz, Operator Collective, Oregon Venture Fund, Essence VC, and Alumni Ventures also participated in the round. So far, the Berkeley, California-based startup has raised $30 million in funding.