European energy customer churn model
Built an AWS lakehouse with 11 data ingestion flows and churn prediction models on SageMaker for an energy company's electricity customers. The model moved from proof of concept to production over about a year and feeds marketer reports on two-month churn probability. EPAM reports over 90% accuracy against a 65–70% contractual expectation.
epam.com · checked 17 Nov 2026 (opens in a new tab)