Precise short-term blood glucose forecasting is critical for managing type 1 diabetes (T1D), enabling timely interventions to prevent complications. Complex deep learning models often rely on intricate architectures, increasing computational demands. We propose an ensemble model integrating XGBoost and LSTM, evaluated on the OhioT1DM dataset. Our preprocessing pipeline employs Kalman Filter smoothing, oversampling of high-glucose events, and feature engineering to extract lagged, rolling, and interaction features. The ensemble achieves an RMSE of 19.89 mg/dL, an MAE of 10.37 mg/dL, and 99.20% of predictions in Clarke Error Grid (CEG) Zones A and B, outperforming individual models and competing with state-of-the-art methods. Its simple architecture and fast inference time of 1.33 s for 3738 samples support real-time deployment on constrained devices such as smartphones. Feature importance and uncertainty estimation (85.58% coverage for 90% prediction intervals) enhance clinical usability, while future work will explore generalization to diverse cohorts and longer prediction horizons.