The present thesis aims to develop a Deep Learning-based Multilayer Perceptron (MLP) model to invert the fundamental-mode Rayleigh wave dispersion curve obtained from the MASW tests to predict the shear wave velocity profile. The traditional inversion technique, such as the local search method, is prone to falling into local extrema, while the global search method offers a rigorous and computationally expensive solution. The majority of existing AI-based inversion methods are either site-specific or have been trained using synthetic datasets with limited variations in subsurface model parameters, which limits the efficacy of the AI model for experimental dispersion data. In the present thesis, we propose two separate MLP models to evaluate the thickness and shear wave velocity from the dispersion data. The MLP models were trained using a dataset generated from a set of random synthetic velocity layer models, accounting for a wide range of realistic subsurface models. The optimal parameters of the MLP network are determined by performing a series of experiments on the training dataset to establish the best MLP architecture. The accuracy of the trained MLP models were tested over the synthetic test dataset and synthetic dataset obtained from the literature. Finally, the MLP predictions were compared with the popular commercial software and published inversion results. The results illustrate that proposed MLP models have great potential to be used as tools for the inversion of Rayleigh waves.