Production ML that moves money.
Before the agent work, there were models. Three systems built for enterprise clients where predictions directly drove decisions — marketing budgets reallocated, customers retained, loans issued. Each one in production.
Past engagements
Selected ML builds
Lifetime Value Forecasting Engine
Built for a €1B-revenue casino
The marketing team had budget to spend but no way to know which regions returned the most value. We built a three-part Python model — retention probability, predicted active days, and average daily revenue — piped into QlikView with a fact-vs-forecast dashboard. Budget reallocation became data-driven.
Credit Card Churn Detection
Trained on card transaction behaviour
Customers rarely announce they're leaving. We found that declining credit card usage was the leading signal. The model segmented cardholders into loyal, at-risk, and departed cohorts — giving the retention team an early-warning list before customers closed their accounts.
Multi-State Credit Scoring
Deployed across multiple US state jurisdictions
Expanding into new US states meant different underwriting regulations per market. We built adapted scorecards for each jurisdiction, letting the microfinance company approve borrowers at scale without running afoul of state-level lending rules.
Have a model to build?
We take a small number of custom ML engagements per quarter. If you have a prediction problem and the data to solve it, let's talk.
Book a free discovery call