In response to the threat of climate change, ambitious targets have been defined at European level for the environmental performance of the aviation industry, namely dramatic reductions in greenhouse gas emissions by 2035 and climate neutrality by 2050. These challenging objectives can only be achieved through radical innovation in aircraft propulsion, including revolutionary engine architectures and propulsion systems based on alternative fuels. However, such changes in engine design imply the ability for aerodynamic simulation tools to explore a larger design space with a higher level of representativeness than current RANS-based CFD solvers are capable of. Scale-Resolving Simulation (SRS) methods, namely Large-Eddy Simulation (LES) and Direct Numerical Simulation (DNS), can significantly contribute to overcome the current barriers. SRS is nowadays becoming applicable to turbomachine flows of industrial interest by means of mature high-order numerical methods in combination with new-generation supercomputers, that offer unprecedented computational power. In this talk, we will present on-going development efforts at Cenaero to build an SRS capability based on this principle, with emphasis on two aspects. The first one is the pursuit of high fidelity, both numerical (with high-order Discontinuous Galerkin method and accurate shock-capturing techniques) and physical (with representative inlet turbulence injection techniques). The second one is the efficient exploitation of modern High-Performance Computing (HPC) resources, with parallel mesh adaptation techniques, statistical convergence estimators and implementation on modern supercomputing architectures. Examples of applications will be given. Despite the possibilities offered by the latest HPC systems, SRS methods remain too costly and specialized to be directly used for design optimization in industrial contexts, which limits their impact on turbomachine technologies. Another path to exploit the benefits of SRS is the use of LES and DNS results to benchmark, calibrate and improve turbulence models, particularly with the help of Machine Learning (ML) techniques. We will show how we employ SRS methods to generate high-fidelity turbulent flow databases and create new ML-based wall models.