Transition-metal systems play key roles in catalysis, photochemistry, and spectroscopy, but their electronic structures are often challenging to describe because of strong electron correlation, spin-state energetics, and dense manifolds of excited states. Multiconfigurational electronic-structure methods provide a reliable framework for studying such systems and are particularly important for the interpretation of X-ray spectroscopic measurements. The overall aim of this project is to advance the computational study of transition-metal electronic structures using multiconfigurational approaches and to develop efficient methodologies that extend their applicability to increasingly complex chemical problems. A central direction of the project is the development of machine-learning-assisted approaches for multiconfigurational calculations of X-ray spectra. High-level treatments of dynamic correlation, such as RASPT2, are essential for obtaining quantitatively accurate excitation energies and spectral features but often represent the computational bottleneck in simulations of core-excited states. We will generate benchmark datasets from large-scale multiconfigurational calculations and use them to train machine-learning models that predict dynamic-correlation contributions from lower-level electronic-structure descriptors. The goal is to establish an interpretable and transferable workflow capable of reproducing near-RASPT2-quality spectral predictions at substantially reduced computational cost. The resulting methodology will facilitate routine calculations of X-ray spectroscopic observables for chemically relevant transition-metal systems. The project will also apply multiconfigurational electronic-structure methods to investigate electronically excited states of molecular transition-metal catalysts relevant to solar-energy conversion and CO₂ reduction. In particular, we will study an iron-based photocatalyst for which recent time-resolved X-ray spectroscopy experiments have revealed ultrafast spin-state changes following photoexcitation. Multiconfigurational calculations will be used to characterize excited-state potential-energy surfaces, electronic-state couplings, spin-orbit interactions, and vibronic effects. These electronic-structure data will further serve as input for quantum-dynamical simulations aimed at understanding the experimentally observed relaxation pathways and their relationship to catalytic performance. Together, these activities combine methodological development and application within a common framework centered on multiconfigurational descriptions of transition-metal electronic structure. The project will generate benchmark datasets, computational workflows, and mechanistic insight into electronically excited transition-metal complexes while strengthening the connection between advanced electronic-structure theory and modern X-ray spectroscopy. The computational workload involves extensive RASSCF/RASPT2 calculations, large sets of core-excited states, and electronic-structure calculations for excited-state models, making access to national high-performance computing resources essential.