Work
These are explorations, not client work. Every company here is fictional and every number is synthetic or drawn from public data. They show how I approach a problem, not results for any real business.
“Northwind Retail” · fictional
The question: can leadership see revenue, orders, average order value, acquisition cost, retention cohorts and marketing return on one page, defined once and trusted by everyone? The exploration builds the dashboard and a short monthly MIS report around it.
“Contoso Play” · fictional
A gradient-boosted churn model set honestly against a simple baseline. On this synthetic data the gain is small (AUC 0.80 against 0.79), which is itself the finding: the value lies less in the score than in understanding the drivers (SHAP) and in where to act (gains curve).
“Acme Learning” · fictional
What does readiness for India's DPDP Act look like for a mid-sized organisation? This exploration scores each processing activity, surfaces the gaps (in this sample, 50% overall and 23 gaps) and works a roadmap back from the date core obligations apply. Not legal advice.
Synthetic data
An invoice-level view of supplier payments by MSME category, with days outstanding measured against the statutory 15- and 45-day limits. The companion field note explains the rules.
Small open-source demos on synthetic data. They show methods only and contain no employer data.