Gender bias in neural machine translation is not a new topic. However, works on reducing such bias from natural language processing tasks are fairly new. In this paper, we have tried to analyze the gender bias that is observed during machine translation, one of the most notable applications of natural language processing, of three of the South Asian languages. We have then tried to reduce the bias using one of the pre-existing methods, specifically by using more data for the biased segment of the population. The purpose of this paper therefore is to help the current machine translation systems reach a better state with more accurate prediction of the subject’s gender, depending less on other factors (mostly occupation), that are often unrelated.