The segmentation model aims to generate an MRI image with added segmentations like edema, enhanced tumor, and tumor core regions based on a pre-treatment MRI image. The segmentation model will be developed based on the 3D full-resolution nnU-Net architecture, a self-adapting framework for medical image segmentation. The model will be trained using the BraTS 2021 dataset, which comprises 369 3D MRI scans across the T1 modality. The dataset includes annotations for Edema, Enhancing Tumor, and Tumor Core. In addition to the standard augmentations that nnU-Net applies to increase the dataset size, Brainstorm will employ new augmentation techniques such as Nine Dot Moving Least Squares (ND-MLD) during preprocessing. This method uses nine control points arranged in a 3 x 3 grid, strategically placed on an image to generate new variations and artificially expand the dataset. The combined Dice loss, which handles class imbalance, and the Cross-Entropy loss, which penalizes pixel-wise misclassification, are specifically chosen to better generalize the training data. To prevent overfitting, nnU-Net's dynamic architecture also continues to adjust factors such as the learning rate and uses regularization techniques like dropout and weight decay. As a proof of concept, Brainstorm has developed the segmentation agent with only the standard augmentation techniques, which resulted in a score of 0.89 IOU. When Brainstorm gets additional funding, we plan on implementing ND-MLD to increase our dataset size by 5x.