<p>The integration of machine learning in the lending industry has not just significantly enhanced efficiency and speed in loan approval processes, but has also raised significant concerns about its potential to propagate existing biases, particularly in the context of racial disparities. Theoretically, the distributional consequences of advanced ML algorithms in loan approval processes can arise from their enhanced ability to uncover complex relationships in historical data, which may reflect historical discrimination. This paper examines methods aiming to reduce the impact of ML algorithms on racial bias in mortgage lending using HMDA data from 2007 to 2017. The analysis involves predicting loan approvals using traditional methods, such as Ordinary Least Squares (OLS), and advanced ML models, including Tree-based models, Support Vector Machines (SVM), and Neural Networks. Furthermore, it introduces a debiasing technique DualFair. Lastly, the paper develops and applies a novel "Multi-Model Bias Mitigation and Consolidation" debiasing extension, along with a novel metric: Fairness-Accuracy score. This enhanced debiasing algorithm significantly reduces bias metrics such as Average Odds Difference (AOD) and Average Weighted Inclusion (AWI), while keeping the accuracy metrics high. The results of this analysis underscore the importance of integrating fair ML algorithms in the financial technology sector to ensure equitable lending practices. By effectively reducing bias, advanced debiasing techniques promote fairness, help to prevent the perpetuation of historical discrimination, and ensure equal access to the mortgage market for all racial groups.</p>