In accordance with the World Health Organization (WHO), depression is the main root cause of illness affecting an astounding 264 million individuals on worldwide level. Depression is an extremely prevalent mental illness health condition and creates difficulties in determining and evaluating the severity of medical condition especially at work place. The rapid pace of contemporary life, coupled with substantial obligations, can have a negative impact on individuals thus enhancing their vulnerability to mental disorders. Therefore, mental health challenges have become more prominent among professional workers. This paper presents a novel approach using machine learning algorithms to predict depression severity based on various mental health indicators containing the data of working professional of both tech and non-tech companies. The comparison has been made for evaluating the performance of traditional algorithms like random forest, SVM, logistic regression, KNN, and Naive Bayes with advanced techniques like gradient boosting, XGBoost, LightGBM, and a voting classifier. The results demonstrate superior predictive accuracy, recall, precision, and F1 score achieved by the voting classifier (accuracy: 45.37%, recall: 45.37%, precision: 56.17%, and F1 score: 42.94%). This study sheds insights on the possible benefits of machine learning in enhancing depression severity prediction, opening avenues for personalized treatment strategies and early intervention programs.