Gender bias in Natural Language Processing (NLP) systems can lead to unfair treatment and discrimination, as these models often reflect societal stereotypes present in their training data. A key obstacle in addressing this issue is the lack of explicit gender information in many textual datasets, making it difficult to measure and mitigate bias effectively. The main goal of this work is to develop novel techniques that accurately identify gender in natural language text and measure and mitigate bias in downstream tasks. This research also extends bias evaluation methods to include non-binary gender identities, promoting inclusivity and ensuring that NLP systems perform equitably across diverse applications. Additionally, the proposed techniques are extendable to measure other types of biases, including those related to race, further broadening the scope of fair and ethical AI development.