A consistent microscopic theory of the atomic nucleus is necessary to explain a wide range of nuclear phenomena and to predict nuclear behavior that cannot be measured experimentally. Over the past decade, theoretical investigations have fundamentally shifted from phenomenological methods to ab initio approaches to understanding nuclei. The availability of high performance computing resources allow for the more reliable, but computationally intensive first principles methods to be tractable.

The No-Core Shell Model (NCSM) is an ab initio approach that has been successful in describing light nuclei. Nuclear properties are obtained through

diagonalization of the Hamiltonian matrix, which is large (dimension > 10^8) and sparse. The computational effort increases exponentially with nucleon number using a 2-body Hamiltonian, and becomes particularly difficult when including 3-body forces, as the matrix becomes less sparse. New algorithms and load balancing techniques are needed to scale this method for larger nuclei (A > 16). We analyze the computational challenges of REDSTICK, a nuclear shell model program, and implement several techniques to scale REDSTICK to investigate high p-shell and low sd-shell nuclei.