Worked on the lanl/singularity-eos repository to enhance the robustness of its Newton-Raphson root-finding algorithm. Addressed a critical bug by refining the convergence check to use absolute value comparison when handling negative ytarget values, which improved the solver’s accuracy and stability in edge cases. This targeted fix reduced the risk of false convergence, leading to more reliable production simulations. The work involved C++ and applied expertise in algorithm optimization and numerical methods. All changes were carefully documented and linked to a specific commit, demonstrating a disciplined approach to code health, traceability, and incremental improvement of scientific computing software.