In silico modelling of medical devices is a very promising strategy for minimizing the requirement for in vitro research and animal testing while boosting device efficiency and cutting development costs. In the current work, in silico models of two commonly used endovascular devices, one interventional and the other implantable, were created. The first device designed was a stentriever for mechanical thrombectomy in the treatment of acute ischaemic stroke. Although stentrievers are routinely utilized in clinics, their true function is not well known. The modelling sought to determine how arterial geometry, thrombus characteristics, and the thrombus' interactions with both the artery wall and the stentriever influence the effectiveness of stentriever thrombectomy. To this goal, we ran finite element simulations of the complete stentriever technique. The modelling included thrombus rupture in order to quantify the risk of embolism. The simulations produce an atlas of failure risk as a function of thrombus composition and artery occlusion site. The modelling findings were then fed into machine learning algorithms to produce a proof-of-concept demonstration of the computational model's use as a real-time patient-specific prediction tool for the likelihood of success of stentriever thrombectomy.