HowDiverse.is began as a side project after a member of the nyala.dev team noticed that a British tech unicorn publicly committed to diversity and inclusion did not reflect that commitment in the faces shown on its own website. The team turned that observation into a lightweight tool that anyone can use to audit the gap between DEI messaging and visible reality, first for the tech sector and then far beyond it.
The analysis workflow is deliberately simple. You open a company website — ideally a careers page or a 'who we are' page — and the tool extracts the images from it, automatically detects the faces, and runs AI-based gender and ethnicity classification on each face. What comes back is a diversity report for that page. Because the underlying model is machine learning rather than a human reviewer, the team states plainly that it regularly reviews outputs manually and warns that the classifier is itself subject to bias.
Results submitted by users are aggregated into public, shareable datasets covering groups such as engineering teams, non-profits, media, government, digital, sports, big tech leadership, venture capital funds, fintech and banking, so the project aims to crowdsource diversity data for organizations around the world. The project positions transparency as the precondition for real accountability, rather than treating diversity metrics as a corporate checkbox exercise.
