As high-performance computing (HPC) shifts toward energy-efficient ARM-based architectures, the scientific community faces a "portability wall": vast archives of legacy scientific kernels hand-optimized in x86 assembly are incompatible with modern hardware. Rewriting these kernels manually is error-prone and costly. This work presents TransISA, a static transpiler designed to automate the translation of x86 assembly source code to semantically equivalent ARMv8-A (AArch64) assembly. Unlike dynamic binary translators that operate as black boxes at runtime, TransISA follows a traditional compiler architecture to ensure transparency and maintainability. It employs a modular transpilation pipeline consisting of: (1) a frontend that lifts x86 assembly (NASM syntax) into architecture-agnostic LLVM Intermediate Representation (IR); (2) a middle-end that performs semantic optimizations, including memory-to-register promotion; and (3) a backend utilizing the LLVM code generator to emit optimized ARM assembly. By targeting the assembly source rather than the binary, TransISA allows developers to inspect, verify, and further tune the output, making it a sustainable tool for long-term software maintenance. Initial results demonstrate that while naive translation introduces instruction bloat due to explicit register simulation, the application of LLVM optimization passes significantly reduces this overhead. This approach provides a scalable pathway for migrating performance-critical legacy code to emerging architectures without losing the benefits of static analysis.