Data Scientist — Telecom Churn Prediction

West African telecom operator, 12M subscribers. Our churn rate is 4.2% monthly and we don't know who's about to leave until after they've ported their number. We need a churn prediction model that identifies high-risk customers 30 days in advance — with enough lead time for our retention team to intervene with an offer. Available data: call records, data usage, recharge patterns, complaints history, geographic location. 24 months of history. Prediction must work at subscriber level, update weekly, and feed directly into our CRM for retention campaign targeting.