The aim of the proposed project is to develop a new hybrid CPU/GPU code for direct simulation Monte Carlo (DSMC) method with dynamic re-partitioning feature and graph theory-based domain decomposition algorithms. DSMC method is a particle method employed for studying rarefied gas dynamics and widely used to study non-equilibrium hypersonic flow around re-entry vehicles at higher altitudes. The proposed code will have efficient load balancing features suitable for extending the dimensionality of the problems from 2D to 3D and aide in simulating hypersonic flows at denser, near-continuum flows at lower altitudes. The new code can also be used by several different branches of flow physics from micro-channel flows to gas-plume expansion flows to vapour deposition techniques to applications in nuclear physics. Further, the graph theory-based domain decomposition algorithms are method independent and can be modified for other Lagrangian simulation techniques.