This NAISS MEDIUM project is grounded in a VR research grant focused on developing emulator-driven and realistic ab initio (first principles) predictions of low-energy quantum structure, binding energies, and electromagnetic transitions in atomic nuclei. Accurate descriptions and predictions of nuclear structure and reactions provide crucial insights into fundamental questions about the limits of stability and origin of visible matter in the Universe. A major limitation in ab initio nuclear physics is the lack of a realistic description of the strong nuclear interaction with quantified uncertainties. Improving our understanding of nuclear interactions requires repeatedly solving the many-nucleon Schrödinger equation using multiple models of nuclear forces grounded in chiral effective field theory. A typical computational statistics analysis will require at least tens of thousands of runs. Unfortunately, solving the nuclear many-body problem is computationally very expensive, even for a single run, typically requiring several thousands of CPU-hours per nucleus and per interaction model. Moreover. The cost of solving the Schrödinger equation (naively) scales exponentially with nucleon number. As such, moving up in mass across the nuclear chart amplifies the computational challenge correspondingly. Recently, polynomial-scaling methods, such as coupled-cluster and similarity renormalization group methods, have significantly increased our reach across the nuclear chart. Nevertheless, such methods remain computationally demanding, typically necessitating leadership-class computing resources for the heaviest nuclei [Nature Physics 18, 1196 (2022)]. This project will use projection-based emulation to address the computational bottleneck of repeatedly solving the Schrödinger equation in the study of nuclear interactions. As demonstrated in some of my recent works [Rev. Mod. Phys. 96, 031002 (2024)], projection-based emulators, i.e., surrogate models that mimic the results of exact simulations, yield extremely accurate approximations (99%) at significantly reduced computational costs. This method leverages a mathematical transformation of the original complex many-body problem into a carefully chosen subspace, using only a limited number of expensive exact solutions to generate fast and accurate predictions across extensive and high-dimensional parameter spaces. This is akin to generating training data for a supervised ML approach. Indeed, this emulation technique shares conceptual similarities with ML methods but remains distinct from common (non-intrusive) approaches such as neural networks or Gaussian processes. Our computational efforts in the NAISS MEDIUM project is focused on generating emulator training-data, i.e., exact wave functions, for a range of nuclei critical for contemporary experimental investigations at radioactive ion beam facilities worldwide. Specifically, we will study selected isotopes of Neon, Oxygen, Magnesium, Calcium, Nickel, and Tin. These nuclei are central to theoretical and experimental programs probing nuclear stability limits and testing the predictive power of ab initio nuclear theory. Once trained, the emulators will efficiently run on Chalmers local single-node computing environments, enabling comprehensive statistical analyses crucial for extracting reliable physical insights. To summarise, this NAISS MEDIUM project is critical for generating training datasets and constructing high-fidelity emulators necessary for my VR research grant.