Skin cancer is a growing global health concern, with millions of new cases reported annually. Early detection is crucial for successful treatment and improved patient outcomes. Deep learning algorithms have shown great promise in skin cancer diagnosis, with the potential for self-monitoring applications that enable timely medical consultation. However, current deep learning systems face limitations due to narrow training datasets and a lack of flexibility in real-world settings. This study focuses on developing a robust deep learning framework with an enhanced optimizer to overcome these challenges. Experiments on ISIC, HAM10000, and a mixed dataset demonstrate superior accuracy and generalization using the proposed approach compared to regular optimizers. The findings highlight the potential of deep learning in proactive healthcare and emphasize the need for further research to improve model accuracy, expand dataset diversity, and enhance user experience for widespread adoption. The study introduces a mixed dataset, an enhanced optimizer, and a web application for early skin cancer detection, aiming to facilitate patient self-monitoring and alleviate the burden on healthcare practitioners .