My research focuses on optimizing data center systems, the backbone of modern society’s essential services. As Dennard scaling ends and data growth surges, datacenters are turning to multi-core parallelism, accelerators, and heterogeneous memory and storage technologies. Yet, this data influx may soon test these accelerators’ scalability. To tackle this, my research advocates software-driven integration of legacy, current, and future heterogeneous computing elements for collaborative data processing. My group is developing cross-layered, scalable, disaggregated operating systems to leverage heterogeneous hardware to design and develop unified, efficient, adaptable, and sustainable data-processing systems.