This project presents a robust machine learning pipeline for forecasting next-day cryptocurrency volatility using an XGBoost model. The solution includes a reproducible data workflow, model training with time-series validation, and an interactive Streamlit application for real-time prediction and risk regime classification.