Mahshook’s work addresses which host–defect combinations produce usable quantum states. A deep-level centre — a point defect whose electronic states sit isolated within the bandgap — can act as a spin qubit, but only a small fraction of possible defect–host pairs satisfy the requirements on level position, charge-state stability, and spin multiplicity at once. The combinatorial space is far too large to search by direct calculation.

He builds predictive models trained on computed and experimental data to identify candidate pairs and to isolate the structural and electronic features that determine whether a defect level falls in the useful range.

Using a composition-only machine learning approach built on ensembles of diverse classifiers, he screened about 45,000 thermodynamically stable compounds to extract shared design rules—such as filled valence shells and low chemical heterogeneity. This screening recovered every experimentally known host (diamond, SiC, ZnO, ZnS) and identified 122 high-confidence candidate materials, including TiO₂, PbWO₄, and layered HfS₂ and ZrS₂, for subsequent first-principles and experimental investigation.