Developed robust solver infrastructure for the ORNL/ReSolve repository, focusing on scalable and portable performance across CPU, CUDA, and HIP backends. Built a Schur Complement Conjugate Gradient (SCCG) solver and a new Matrix class supporting sparse matrices, with comprehensive unit and cross-backend tests to ensure reliability. Emphasized explicit memory management in C++ by removing automatic allocations for matrix and vector operations, improving control and robustness. Implemented heap-allocated internal vectors to prevent stack overflows and aligned backend-specific handlers for matrix-vector operations. Enhanced test coverage and addressed backend consistency, demonstrating depth in algorithm design, memory management, and cross-platform optimization.