Integrated material system modeling includes material law discovery, material parameters identification, and forward (multiscale) simulation. In traditional material system modeling, these 3 moduli are segregated into different forward and inverse solvers. A seamless integrated solution is proposed using the idea of differential simulation. In the material law discovery module, novel sparsity-promoting techniques such dimensionless learning can be used to discover physics-constrained material law [1]. In the material parameter identification module, the inverse problem can be efficiently solved using the adjoint method, where the Jacobian matrix of the forward solver is obtained using automatic differentiation. In the forward (multiscale) simulation module, C-HiDeNN enables higher order smoothness and accuracy. Fast multiscale modeling is achieved by using kernel learning method as a surrogate model for microscale analysis.