GPU Virtualization

Novel orchestration and scheduling methods for API remoting-based GPU virtualization on Kubernetes clusters, aimed at democratizing access to costly GPU compute and memory resources.

ai-systems in-progress
ccudakubernetesdockerpython

GPU Virtualization

This project develops orchestration and scheduling methods for API remoting-based GPU virtualization techniques to improve GPU cluster utilization. The goal is to democratize access to costly GPU compute and memory resources across shared infrastructure.

Key Results

  • Deploying and evaluating state-of-the-art API remoting-based GPU virtualization on Kubernetes clusters
  • Novel scheduling methods designed to maximize GPU utilization across multi-tenant environments
  • Active research at Florida International University (Postdoctoral Associate, Sep 2025 – present)