The interview combined past‑project experience with technical questions on predictive methods, algorithms, and their applications, focusing on model choices, evaluation metrics, and business impact from my resume.“In one project we predicted customer churn using historical usage and billing data. We treated it as a binary classification problem and used XGBoost because it handles mixed‑type features well and gives feature importance. We did a time‑based train/validation split, tuned depth and learning rate via cross‑validation, and evaluated on AUC and F1. Feature importance showed that billing‑related features were the top drivers, and the model helped the product team prioritize retention campaigns on high‑risk segments.