Application of Kinetic gas theory in computational fluid dynamics can bridge the gap between molecular and macroscopic scale models. The physics informed neural networks can then learn the underlying statistical mechanics and mimic the results within considerably shorter time frames, and help us understand the interactions between scales.