Effective recruitment planning is crucial for the success of clinical trials, as inadequate recruitment is a leading cause of trial discontinuation. Such setbacks not only cause delays and incur additional costs but also result in lost market opportunities. In response to this challenge, the development of a predictive tool for recruitment planning has garnered significant interest. Our work focuses on the intricacies of screening and enrollment forecasting for trials involving various disease subtypes or biomarker-defined subgroups within a multi-regional context. We employ a hierarchical semi-parametric Bayesian approach to model recruitment dynamics, which allows for the accommodation of heterogeneity across patient populations attributable to regional or subgroup distinctions. To support ongoing recruitment efforts, we propose a dynamic prediction framework that offers continuous guidance for recruitment planning. This innovative tool, grounded in model-based and data-driven methodologies, demonstrates notable flexibility and precision in projecting enrollment figures. Moreover, it efficiently identifies potential future risks, thereby facilitating informed decision-making in the management of clinical trials.