Bangladesh’s coastal zone is widely recognized as one of the most hazard-prone regions of the world. Within this vulnerable belt, Lakshmipur District stands for its long, highly exposed coastline and its recurrent experience of hydrometeorological hazards, including frequent floods, tidal surges, river overflows, and intense monsoon rainfall. The district’s flat topography and geographic setting further amplify its susceptibility to both seasonal and extreme flood events. Notably, two consecutive major floods in 2024 and 2025 affected approximately 600,000 and 50,000 people, respectively, with the 2024 event alone causing crop losses exceeding BDT 2.27 billion. In response to these escalating risks, this study develops a Geospatial Artificial Intelligence (GeoAI)-based flood risk mapping framework that integrates multi-source satellite Earth observation data and social datasets to map flood extents for 2024 and 2025 and to predict flood risk for Lakshmipur District. Flood extents were derived using Sentinel-1 Synthetic Aperture Radar (SAR) imagery, while flood risk prediction employed a Random Forest classifier trained on elevation, slope, distance to rivers, precipitation, population, cropland, and built-up indices. The resulting probabilistic flood risk was classified into five categories ranging from very low to very high risk, with population and cropland exposure quantified at the union level. Model validation using 2025 flood data achieved an overall accuracy of 85.9% and a Cohen’s Kappa of 0.71, demonstrating strong predictive performance. The results highlight critical flood-risk and exposure hotspots, underscoring the utility of GeoAI-based approaches for enabling timely, efficient, and actionable flood risk assessments, particularly for vulnerable communities where early risk detection through mapping and modeling is essential for proactive disaster planning and targeted mitigation for natural disaster management.