With the advent of modern numerical simulation tools, virtual human twins are becoming a widely used tool in biomedical research, making it possible to study complex physiological processes in a safe and controlled way. By using physics-based models of the human body, these digital replicas can reproduce the many interactions behind organ function, improving the understanding of diseases and opening the door to more personalised treatment strategies, without the need for invasive experiments. In this talk, we start by introducing a high-fidelity framework for cardiac simulations including electrophysiology, tissue mechanics, and haemodynamics across multiple spatial and temporal scales. The approach is based on Alya, a high-performance finite element solver developed at the Barcelona Supercomputing Center and at ELEM Biotech, which has shown significant success in accurate simulations of cardiac dynamics under both healthy and pathological conditions. We conclude by discussing how artificial intelligence can be used to construct efficient surrogate models, strongly accelerating the highly expensive Alya-based simulations.