Our Story

We built what we needed at 3 AM.

In early 2023, Kevin Nakamura was leading ML platform engineering at a Seattle-based fintech, maintaining 22 production models across four product lines — credit scoring, fraud detection, churn prediction, and loan underwriting. In Q4 review, the compliance team submitted a formal request: document why the credit-risk model had been retrained three times in the past quarter, including who authorized each retrain and what triggered it.

The answer took a week. W&B had the experiment run metrics. MLflow had the artifact lineage. But neither had the approval chain, the policy threshold that triggered the retrain, or an immutable record of who reviewed what and when. The audit evidence was assembled from Confluence, Slack exports, and three separate retrospective documents — none of which were designed to be tamper-evident.

That week, Kevin and Yuki Tanaka sketched the first architecture of what became Inferpathio: governance as a versioned policy layer that sits above your experiment tracker and artifact registry, not inside them. Inferpathio was incorporated in September 2023. It has been bootstrapped since day one — no outside capital, no investor timelines.