Permanent magnets (PMs) are essential components of energy-efficient technologies, including electric vehicles and wind-power generators. State-of-the-art high-performance magnets, however, rely strongly on critical rare-earth (RE) elements. Computational materials design can accelerate the search for RE-free and RE-lean alternatives, but several key magnetic properties remain difficult to evaluate on the scale required for systematic materials screening. In particular, accurate calculations of magnetocrystalline anisotropy energy (MAE) require high numerical precision and constitute a major computational bottleneck. The proposed project combines first-principles electronic-structure calculations with data-driven approaches and targeted investigations of promising permanent-magnet families. A central activity will be the continuation and extension of an ongoing six-month AI4TekNat-funded project investigating whether information contained in the electronic density of states (DOS) can be used to predict MAE. During the current allocation period, DOS calculations have been performed for a large set of materials and selected MAE values have been recalculated to establish a consistent reference dataset. The first iteration of the database for machine-learning analysis is now approaching completion. Additional calculations will be required as the ML analysis identifies missing regions of materials space, anomalous data and systems requiring higher-accuracy verification. Beyond the immediate funded project, the resulting dataset will also be used in collaboration with researchers specialising in signal analysis to explore alternative methods for identifying physically meaningful features in electronic spectra. A second activity concerns RE-lean Ce-based 2:17 permanent magnets. Structural and magnetic data generated during the previous allocation contribute to the MINT-ATLAS database project, and a joint publication with several collaborators is currently in preparation. Further targeted calculations will be performed where required to complete and validate this dataset. Finally, a new research direction will be initiated within the recently funded European SPEED-Mat project. Preliminary first-principles studies will focus on representative 1:12-type RE-Fe-based permanent magnets. These compounds offer a promising route towards reducing critical-element content while retaining favourable intrinsic magnetic properties, but their stability and magnetic performance are highly sensitive to chemical composition. Initial calculations will establish reliable computational settings and provide theoretical guidance for subsequent materials optimisation within the European project. The project will primarily employ VASP for structural, electronic and relativistic MAE calculations, complemented by RSPt for detailed magnetic-property calculations and, where required, SPRKKR and UppASD.