In this integrated project we propose the concept of a quantum virtual machine (QVM) to study comprehensively the potential of quantum advantage in and beyond noisy intermediate-scale quantum (NISQ) era. A QVM consists of three layers, separated by their level of abstraction. At the top we envision a quantum application layer, where we plan to perform quantum simulations of topological quantum dynamics and quantum field theories. At the intermediate level, we envision a quantum architecture layer, where we explore quantum-classical hybrid architecture inspired by the interplay between quantum circuit, tensor network, and machine learning. At the bottom, we envision a quantum middleware layer, in which we explore the optimization and characterization of NISQ processors in a device independent fashion. The layer will also serve as a bridge to the underlying quantum firmware layer, which performs device-dependent controls. We expect that the project will provide a unique opportunity to integrate the quantum software stack with domestic and international NISQ processors.