The THRONE (THRombosis Operator NEtwork) project aims to develop an innovative multi-scale modeling framework for simulating complex cardiovascular phenomena, specifically focusing on high-shear induced thromboembolic events. Thrombosis, haemolysis, and embolism are critical pathophysiological processes that challenge state-of-the-art modeling approaches due to their multi-scale nature—spanning from the molecular level (nanometers and nanoseconds) to the macroscopic level of clinical data (centimeters and seconds). Addressing these phenomena is key to advancing our understanding of cardiovascular diseases, the leading cause of death worldwide. The primary objective of the THRONE project is to integrate molecular dynamics simulations, computational fluid dynamics (CFD), and neural networks to achieve a complete, efficient description of shear-induced thrombosis and related processes. The project leverages neural operator networks, specifically DeepONet, to couple microscopic platelet dynamics simulations with macroscopic blood flow simulations. This approach efficiently bridges scales, enabling real-time predictions of platelet activation, adhesion, and aggregation under various blood flow conditions. At the microscale, the behavior of individual blood components (e.g., platelets, red blood cells) is modeled using particle-based methods, capturing their detailed response to shear stress and other flow conditions. These particle-level simulations, while highly accurate, are computationally expensive. To address this, neural operators are trained on the results of these microscale simulations, learning to predict platelet behavior as a function of local blood flow conditions. Once trained, these neural operators can be integrated into macroscopic CFD simulations, drastically reducing computational costs while preserving accuracy. This eliminates the need to run full-scale particle simulations for each time step in the macroscopic flow model, making the THRONE framework highly efficient and scalable. At the macroscale, blood is treated as a multiphase fluid, with individual blood components represented as solid particles transported within the flow. These CFD simulations are carried out using spectral element method-based solvers (e.g., NEK5000 or Neko), providing high resolution of complex blood flow dynamics in both patient-specific geometries and medical devices. The use of Large Eddy Simulations (LES) further allows the model to resolve turbulent scales without excessive computational cost, a critical feature for simulating blood flow in high-shear environments, such as stenotic arteries or mechanical circulatory support devices. Given the computational intensity of the neural operator training, GPU resources are essential to accelerate the training process. Access to high-performance GPU clusters will enable efficient model development and deployment, ensuring the success of this groundbreaking initiative in cardiovascular modeling.