Extreme Multi-Label Classification (XMLC) deals with classification problems where the number of labels can reach millions, and each instance may be associated with a small subset of relevant labels. XMLC is used in a wide variety of domains, including recommendation systems, web categorization, document tagging, product search, and personalized advertisements. The primary challenge lies in efficiently handling the enormous label space and ensuring scalability, precision, and performance when only a few relevant labels are associated with each instance.Extreme Multi-Label Classification (XMLC) open up opportunities to work on cutting-edge challenges across domains such as recommendation systems, document categorization, and personalized services. These projects explore innovative methods like deep learning architectures, attention mechanisms, graph-based approaches, few-shot learning, and adversarial training to handle the complexity and scale of XMLC tasks. Each project addresses unique research challenges, aiming to improve both performance and scalability in real-world applications.