How this portfolio is built

Every Material Signal project is two things: an analytical engine written in Python, and a set of pages that render what that engine produced.

The engine lives in analytics/. It generates a synthetic dataset with its problems deliberately seeded, runs the analysis, and writes the result to disk as JSON. The website reads those files. No analysis happens when you load a page.

That split is deliberate. The accuracy figures on each project page are measured against a ground-truth record of exactly which records were made wrong and how, rather than asserted. And the same engine that produced them is the one a client engagement would start from.