About
Bias in data is measurable. So remove it, and prove it.
Rosa exists because encoded bias is not a vibe or a policy gap: it is a measurable, removable property of a dataset. Most of the industry measures it and writes reports. Rosa removes it and emits the proof.
The technology, briefly
Rosa is built on the Fair Adversarial Network (FAN): a discriminator repeatedly tries to recover a protected attribute from the data, and the network is trained until it cannot. That adversarial principle is why Rosa finds proxy bias without anyone labelling the proxies.
It ships as a working service: a customer portal, an asynchronous REST API, and an eight-tool MCP (Model Context Protocol) server for AI-native toolchains, with an immutable Run Manifest emitted for every job. Evidence-first is a design decision, not a feature request.
The FAN technology has years of research and validation behind it. It was originally developed by illumr Ltd in London, where the Demonstrating Rosa white paper (PDF) was published. The Rosa intellectual property has since been acquired and is now operated and developed independently.
Where Rosa is today
Live, free to try, and deliberately honest about scope: univariate debiasing (one protected attribute per run), dataset-specific residuals, published validation, and a free Diagnose tier for regulators with no end date. The product earns the trust; we do not borrow it from logo walls.
Remove the bias from your data. Keep the evidence.
Free to try. Free for regulators, with no end date. No credit card.