Improving road safety in increasingly complex and multimodal transport networks requires analytical tools capable of capturing the multifaceted interactions between roadway geometry, driver behavior, and environmental conditions. Traditional statistical approaches though valuable often struggle to model nonlinear relationships and hidden dependencies that influence crash likelihood and severity across diverse transportation contexts. Recent advancements in machine learning offer a more adaptive and data-rich pathway for predicting crash severity by learning from large, heterogeneous datasets that integrate roadway design attributes, behavioral indicators, traffic flow characteristics, and real-time environmental factors. Machine learning algorithms such as gradient boosting machines, random forests, support vector machines, and deep neural networks excel in uncovering nonlinear patterns and high-order interactions that traditional models may overlook. When enriched with high-resolution roadway geometry data, including curvature, lane width, grade, and intersection design, these models can better capture how infrastructure directly shapes collision risk. Integrating driver behavior indicators, such as speeding patterns, distraction proxies, acceleration variability, and compliance with traffic control devices, further strengthens predictive accuracy by reflecting human factors that often precede severe crashes. Environmental elements weather, visibility, lighting conditions, and seasonal effects introduce additional variability that machine learning techniques can incorporate dynamically. In multimodal networks, where cars, cyclists, pedestrians, and transit vehicles converge, machine learning provides nuanced insights into exposure risk and conflict points by fusing sensor data, geospatial information, and historical crash records. These predictive capabilities enable transportation agencies to prioritize high-risk corridors, evaluate countermeasure effectiveness, and design targeted interventions that enhance safety across all modes of travel. By bridging infrastructure, behavior, and environmental analytics, machine learning-based crash severity prediction supports proactive, evidence-driven road safety management and more resilient urban mobility planning.