Increasing Robustness of QPU Callibration with Bayesian Optimization

Quantum processing units (QPUs) require frequent re-tuning across devices and thermal cycles, a process that is labor-intensive and poorly served by inflexible directed acyclic graph based automation. Because QPU data is scarce and costly, we present a data-efficient approach that augments existing calibration pipelines with Gaussian Process based Bayesian optimization to obtain an optimal exploration/exploitation trade off. We demonstrate this by optimizing gate calibrations on a superconducting QPU, where Gaussian Process models efficiently converge to high-fidelity gate parameters, reducing expert intervention and enabling more flexible, scalable QPU operation.