Improving the scalability of electronic structure codes on emerging supercomputers is a significant challenge for computational scientists in the exascale era. We developed a PETSc/SLEPc based Shift-and-Invert Parallel spectral transformations eigensolver (SIPs) to confront one of the scalability bottlenecks of quantum chemistry codes. We benchmarked the scalability of the SIPs on Mira, Theta and Cori and demonstrated superior performance compared to dense eigensolvers (ELPA, Elemental) without any degradation in the accuracy of the numerical results. Recently, we integrated the SIPs into the SIESTA ab initio molecular dynamics package and the Electronic Structure Infrastructure (ELSI) library. As we move more towards data-driven science, another important challenge in quantum chemistry is to generate high-quality data that can be fed into machine learning algorithms. Supported by an exascale computing project, we developed a code, QTC, to automate and parallelize quantum chemistry calculations with the goal of obtaining predictive thermo-chemical parameters for gas-phase chemical kinetics. I’ll present our initial results computed with QTC for the thermochemistry of butane combustion mechanism.