Brain tumors pose significant health challenges, so diagnosis accurately and early in the stage is very essential. In this paper, we have proposed a multilayered neural network model for classifying brain tumors in MRI images. To design the proposed model, a huge dataset has been used, advanced pre-processing techniques have been applied, and a very sophisticated convolutional neural network architecture has been developed that will automatically extract features and classify them with high accuracy. Key innovations here are the inclusion of data augmentation to improve model robustness and the use of fine tuned hyper parameters to improve accuracy. The evaluation metrics used included accuracy, precision, recall, F1 score, and ROCAUC. It can be observed from the results that the proposed technique performed better than all the traditional and other available deep existing methods. Therefore, this will enable the medical fraternity with a reliable tool for better diagnoses and potentially better patient outcomes. The dataset will be increased, and then further fine-tuning of the model will take place for wider applicability with improved performance.