Large-scale industrial problems are typically very complex and difficult to solve by even the state-of-the-art sequential optimization algorithms, and therefore, there is a need to accelerate these algorithms by using modern parallel computing devices like graphics processing units (GPUs) or coprocessors Intel Xeon Phi. Our parallel algorithms for Resource Constrained Project Scheduling and Nurse Rerostering problem (see our RCPSP and NRRP repositories) indicate that combinatorial algorithms can be accelerated up to two orders of magnitude in comparison to sequential processor code. Our research in this domain has been supported by an Nvidia academic grant.