This project is to develop data-driven machine learning-based approximate models, which can replace the time consuming CFD simulations. The proposed work aims to provide proof of concept, methodology to develop machine learning models and its applicability in tasks such as flow flied prediction over variable geometries and geometrical optimization.