Public radiotherapy systems in resource-limited and lower–middle-income countries face critical challenges in efficiently scheduling treatments while ensuring equitable access. These systems often operate with limited technological infrastructure and a small number of treatment machines, leading to capacity saturation and long waiting times. This study focuses on a public oncology center under severe resource constraints, using real-world data to design and evaluate improved scheduling strategies. We present a hybrid decision-support framework that integrates patient segmentation, discrete-event simulation (DES), and online optimization to support treatment planning. Patients are grouped into archetypes based on treatment duration using Jenks Natural Breaks, with model selection guided by the Akaike Information Criterion. These archetypes are used to generate synthetic demand within a DES environment, which enables patient-level tracking to monitor treatment trajectories and identify delays in individual cases. A linear online optimization model assigns each patient to a LINAC and dynamically determines the treatment start date, taking into account indivisible sessions, machine capacity per day, and administrative constraints. The objective function minimizes start time and includes a secondary term to promote balanced use of available resources. The framework allows for comparative analysis of scheduling strategies: the optimized policy is benchmarked against a heuristic baseline that assigns patients to the earliest available slot. Simulation results show improvements in system performance, including more balanced capacity utilization, fewer periods of high saturation, and reductions in patient balking. The proposed framework is flexible, data-driven, and transferable to other healthcare systems with structural limitations, offering valuable support for operational decision-making in radiotherapy services.