Developed a multi-modal data augmentation algorithm that generates new image-text pairsto enhance data efficiency and model performance in visual-language representation learning tasks.