Skin cancer, with its constantly growing incidence, demands methods for early diagnosis, which is essential for successful treatment. Aiming at this objective, this study proposes a web tool based on Deep Learning. The study encompassed the selection and organization of a dataset, the implementation of pre-trained neural networks, the application of regularization techniques, and the development of the web application. The tool uses a convolutional neural network (CNN) trained on the HAM10000 dataset of dermoscopic images to identify suspicious lesions and assist health professionals in assessing the probability of malignancy. Among the tested networks, ResNet152V2 stood out, achieving an accuracy of 99.0% in the test set. The web interface, developed in Flask, allows the upload of photos and the obtaining of results in 1.17 1 Graduando de Sistemas de Informação pela Unichristus; jonasherminio@live.com; 2 Mestre em inteligência computacional pelo centro de informática da UFPE; rodrigo.valentim@unichristus.edu.br; 4 s on average. This study is configured as a contribution to skin cancer diagnostic studies using Deep Learning.