MacLellan, Benjamin2025-11-182025-11-182025-11-182025-11-17https://hdl.handle.net/10012/22630Precisely measuring the natural world around us underpins new scientific discoveries and technological innovation. Quantum sensors, which harness quantum effects such as superposition and entanglement, represent the frontier of precision measurement and are capable of surpassing conventional limits on measurement sensitivity, precision, and resolution. Such instruments have myriad applications in, e.g., astronomical observations, biological imaging, material science, and geophysical surveys, among many others, and provide new opportunities in the search for new physics, including in gravitational wave detection, searches for dark matter and physics beyond the Standard Model, and probing many-body phenomena such as superconductivity. In recent years, artificial intelligence and machine learning have emerged as a promising paradigm for quantum physics, providing computational tools to extract insights from large scientific datasets, discover structure in complex models, and automate scientific processes. In this thesis, we present novel numerical techniques for the study, design, and implementation of quantum sensing protocols by leveraging machine learning and optimization techniques. First, we propose and demonstrate an end-to-end variational quantum sensing framework, in which parameterized quantum circuits and neural networks form adaptive, trainable ansätze for the quantum dynamics and estimator, respectively. Extending this machine-learning design perspective, we study quantum-enhanced very-long baseline imaging, which uses entanglement distributed through a quantum network to increase the achievable angular resolution of optical telescope arrays. We develop differentiable simulation and optimization techniques to identify optimal resource states and measurements in realistic regimes. Next, we propose and demonstrate a simulation-based inference technique for quantum sensing protocols, which maps observed data to estimated values without the need for explicit likelihood functions. Finally, we investigate the generation of quantum graph states using hybrid photon-emitter platforms, and present a framework for optimizing the generation of large, noise-robust entangled probe states for quantum sensing protocols.enquantum sensingquantum informationmachine learningmetrologyautomated designvariational methodsNATURAL SCIENCES::Physics::Other physics::Computational physicsDoctoral Thesis