This project explores the use of machine learning to forecast hourly cryptocurrency volatility for Bitcoin (BTC) and Ethereum (ETH). Using a 14-month dataset from Yahoo Finance, the project engineers 23 structural features, including historical lags, moving averages, and intra-hour measures like Garman-Klass volatility. It implements three distinct machine learning architectures to model non-linear market dynamics: Random Forest, Long Short-Term Memory (LSTM) networks, and XGBoost. These models are evaluated against a naive benchmark—which simply assumes the next hour's volatility equals the current hour's. The results reveal that XGBoost is the top-performing model, achieving approximately a 13% to 15% reduction in Root Mean Squared Error (RMSE) over the baseline.