Selected work.

An end-to-end machine learning project for identifying customers at risk of churn, with a focus on reproducibility, experiment tracking, testing and production-oriented workflows.

A forecasting system designed to predict operational demand and service-related metrics, combining historical patterns, calendar effects and machine learning to support planning and resource allocation.

An operational analytics project that transforms service data into dashboards designed to understand demand, waiting times, SLA performance and operational bottlenecks.

Behavioural clustering that turns raw transaction data into segments a commercial team can act on.

Platform for the full model lifecycle: training, evaluation, registry, deployment, monitoring and automated retraining.

Language models applied to unstructured text: extracting structure, meaning and relations from documents and conversations.