Pulmonary hypertension (PH) is defined as a mean pulmonary arterial pressure (mPAP) of 20 mm Hg or greater at rest, affecting approximately 1% of the global population. While the diagnosis of PH primarily relies on assessing the hemodynamics of pulmonary circulation, and modalities such as 4D-MRI and echocardiography aid in the diagnosis, it is definitively confirmed only through invasive right-sided heart catheterization. The challenges of measuring arterial pressures non-invasively have encouraged the use of computational fluid dynamics (CFD) models and computer simulations. However, despite their predictive power, these computational models have not been widely adopted in clinical practice due to high computational costs and complex deployment procedures. In this study, we introduce a machine learning (ML) approach that circumvents the time-intensive and costly CFD methods to predict pulmonary artery differential pressures. A Multi-layer Feed-Forward Neural Network (MFNN) was trained using anatomical geometry and inlet velocity, alongside pulmonary vascular resistance modeled by the 3-element Windkessel formula. The ML model was tested on a dataset not previously encountered during training, demonstrating remarkable accuracy with a minimum R2 of 92. We aim to integrate this pipeline with imaging modalities to develop patient-specific pressure gradients and potentially reduce reliance on invasive right-sided heart catheterization in clinical settings.