Segmentation of brain magnetic resonance imaging is essential for precise diagnosis, effective treatment planning, and monitoring of neurological disorders. However, low and middle income countries often face significant limitations due to resource constraints and the low quality of imaging. To address these challenges, we propose a hybrid ensemble model combining two advanced segmentation architectures, GLIMS and MedNeXt. By utilizing transfer learning from high-quality datasets, comprehensive fine-tuning, and ensemble fusion techniques, our approach achieves superior performance in segmenting tumors under low-quality imaging conditions. Experimental validation using the BraTS-SSA dataset highlights improvements in accuracy and robustness, positioning this approach as a clinically viable solution for enhancing diagnostic accuracy in resource-limited settings. https://github.com/AliAZ98/GLIMS-MedNeXt .