As I worked on increasingly large engineering simulations, I realized that achieving realistic time-to-solution requires far more than accurate physics models alone. Scientific computing frameworks such as PETSc and MFEM exposed me to scalable solvers, distributed computing, and high performance numerical infrastructures.
These ecosystems shaped my understanding of HPC-oriented software design, scalable algorithms, matrix-free methods, and adaptive parallel computing. They strongly influenced my long-term goal of reducing computational cost while preserving simulation fidelity.