Alzheimer's dementia remains a serious global health challenge, particularly in resource-limited countries where early and accurate diagnosis is crucial to reducing morbidity and mortality rates. Despite advances in medical imaging and diagnostic tools, early detection of Alzheimer’s remains a complex and resource-intensive task for healthcare systems worldwide. This study leverages the power of machine learning, specifically Convolutional Neural Networks (CNN), to develop a reliable model for detecting the severity of dementia using brain MRI images. The dataset used consists of four main categories: Non-Demented, Very Mild Demented, Mild Demented, and Moderate Demented, with a total of 1,561 images obtained from Kaggle. The model was trained using Vertex AI on Google Cloud, which automatically optimized model parameters through AutoML and Hyperparameter Tuning. Techniques such as image segmentation and feature extraction were applied to enhance model accuracy. The results show that this CNN model achieved a precision rate of 93.5% for the Non-Demented category, with classification accuracy consistently between 92% and 93% for various other levels of dementia severity. These findings underscore the potential of machine learning, particularly CNN, in significantly improving dementia detection accuracy even in resource-constrained settings. By utilizing advanced techniques such as image segmentation, feature extraction, and CNN-based automated classification, this model offers a promising solution for real-time dementia diagnosis. The scalability and adaptability of the model built using Vertex AI allow for broader applications in global clinical scenarios, supporting public health efforts to reduce the burden of Alzheimer's disease. While challenges regarding data sensitivity and computational resources are acknowledged, the model’s potential to improve early diagnosis and patient outcomes is highly significant.