In particle-laden flows, Stokes flow is observed at small length scales, where inertial forces are negligible. It involves all types of liquid and gas suspensions, where the size of the particle is smaller than the Kolmogorov length scale, which dictates the size of the smallest turbulent structures in the fluid flow. In dense suspensions, the drag force on particles becomes increasingly influenced by the proximity of neighboring particles. This effect is more pronounced when viscous forces dominate the flow, as is typical in the Stokes flow regime. In the past, probabilistic approaches were developed to simulate the particle drag force in dense flows. Recently, machine learning-based deterministic approaches have emerged, driven by the rapid growth in the popularity of artificial intelligence. While such models have been developed, they have yet to be fully implemented. In present study, the boundary element method (BEM) was employed \cite{Strakl} to compute particle forces in particle-laden Stokes flow. The resulting data was used to train a feed forward neural network (FFNN) model for predicting particle forces, based on the positions of neighbor particles. A total of 14 000 BEM simulations, with varying particle volume fractions, were conducted to generate the training dataset. The particle volume fraction is represented by the actual inter-particle distance. The performance of the model was tested on an independent set of input parameters, on a wide range of particle volume fractions, 1e-8 < φ < 1e-3. The trained FFNN model was implemented in OpenFOAM v11 and used to simulate the particle-laden flow.