TOWARDS ACCELERATION OF QUANTUM-CLASSICAL ORCHESTRATION

Quantum computing is expected to act as a hardware accelerator within future computing infrastructures, complementing classical processors in the same way GPUs accelerate specific workloads today. While algorithms such as Shor’s and Grover’s suggest potential advantages, the practical use of quantum devices will occur within hybrid architectures where classical and quantum resources coexist. Artificial intelligence illustrates this balance: classical methods, such as Transformers and their scalable variants, remain dominant, while quantum computing may only contribute to selected subroutines. At the same time, AI methods are increasingly applied to quantum computing, supporting calibration, control, and error mitigation of near-term devices. This two-way relationship underlines the need for hybrid solutions and for efficient orchestration of heterogeneous resources. This thesis investigates the design and implementation of hybrid classical-quantum solutions, with emphasis on their real-time orchestration within heterogeneous infrastructures. The work has been carried out within the Qibo environment, a fully open-source framework that spans the entire quantum computing stack: from high-level interfaces for algorithm design, to efficient simulation backends on classical hardware, to integration with quantum devices down to the hardware and pulse level on self-hosted systems. This modularity makes Qibo a suitable platform to explore orchestration strategies across classical and quantum resources. Within this framework, we introduce algorithmic and software contributions that accelerate hybrid workloads. We present general-purpose open-source libraries, such as Qiboml for hybrid quantum-classical machine learning, and mpstab for a hybrid stabilizer-tensor network representation of quantum states. By preserving the familiar interfaces of classical machine learning frameworks, Qiboml enables the design of hybrid architectures and their execution on both simulators and quantum hardware. We also develop algorithmic solutions for real-time quantum error mitigation, implemented in Qiboml and relying on Qibo. We further discuss applications, including multi-variable integration, probability density estimation, and ground-state search, as well as an application in which quantum systems operate as precision sensors for fundamental physics. These serve as test cases to validate the proposed approaches and to provide practical examples of how the heterogeneous components of a future quantum-classical infrastructure can be effectively orchestrated.