Low-lying coastal regions of Bangladesh are extremely susceptible to cyclones. With gusts exceeding 110 km/h, Cyclone Remal (2024) caused extensive flooding and wind damage. In order to evaluate wind-induced cyclone damage, this study suggests an integrated framework that combines numerical modeling, machine learning (ML), and remote sensing. Meteorological and hydrodynamic models were used to simulate wind fields (intensity and duration). Damage severity was classified using simulated wind variables by supervised machine learning algorithms (Random Forest, XGBoost) that were trained on ground damage data. Damage indicators such as flooding, vegetation loss, and structure destruction were examined in multi-temporal optical (Sentinel-2, Landsat-8) and radar (Sentinel-1 SAR) satellite images. To measure impacts, change detection algorithms compared imagery taken before and after the cyclone. Improved detection under cloud cover was achieved by combining SAR and optical data. The findings indicate a high degree of agreement between the damage observed and the damage predicted by machine learning (Pearson's r > 0.8). The model's high-risk areas matched damage hotspots found by remote sensing. Severity was affected by exposure, housing type, and landscape characteristics. In cyclone-prone areas, the integrated approach can direct emergency response and long-term risk reduction while facilitating quick post-event mapping.