The scale of urbanization and industrialization is expanding rapidly, which has led to great difficulties in waste treatment and has also led to a substantial adverse effect on the environment and public health. In this study, a prototype of a smart trash can is introduced where automation, embedded technology, and machine learning (ML) algorithms are combined to modify the way waste management is interpreted together. The contraption uses a series of Arduino Uno, ultrasonic sensors, and a mini servo motor for touch-free operation and opens the lid of the bin when it detects trash. It also sorts waste into dryness and wetness by measuring soil moisture using a moisture sensor that gives real-time feedback to the user with the LCD and buzzer. To foster the bin’s intelligence, the ML algorithm was adapted to add the feature of when the bin will be full. A private dataset with synthetic data based on real-world observations is used and balanced with SMOTEENN. To predict fullness, different ML models such as Random Forest (RF), Logistic regression (LR), XGBoost, LGBM, GBM and CatBoost were applied. The model performance was further improved using the ensemble technique, hard voting, combining XGBoost and CatBoost, achieving an accuracy of 98.82%. More metrics such as precision, recall, f1 score and the ROC-AUC curve were generated to evaluate the effectiveness of the model.