Currently, the majority of people in Bangladesh experience cyberbullying, particularly online racism and body shaming. It is a pressing social issue that how to prevent this bullying practice from spreading widely. So, this research tries to develop a dataset and web prototypes that identify racism and body shaming in online. 1350 Bangla sentences were collected from a Facebook post and individual users via a Google form. The sentences were labelled with the terms ‘Racism’ and ‘Body Shaming’ classes. The dataset was divided into 80:20 proportions for training and testing purposes and the machine learning classifier Support Vector Machine (SVM) was applied to achieve better accuracy in percentage 91 and precision 97.50, Recall 83.87 and F1-Score 90.17. After developing the model, a web-based system was developed as a social media prototype using Flask and Bootstrap 5, which allows users to enter Bangla text and then send it to the model for detection. The Flask application handles these incoming requests, and if a sentence has racist or body-shaming meaning, then the system shows a message. In the future, this system can be integrated on a large scale on social media to bring down bullying practices on skin color and body shape and positively impact the internet.