Worked on the InsightSoftwareConsortium/ITK repository to enhance the reproducibility of optimization workflows by addressing non-determinism in multi-start optimization routines. Using C++ and leveraging expertise in optimization algorithms and software engineering, implemented a deterministic approach by resetting the learning rate for each optimization run and introducing an InitialLearningRate parameter. Developed mechanisms to store and restore the learning rate at the start of each optimization, ensuring consistent behavior across repeated runs. Added a dedicated unit test to verify that identical starting parameters yield reproducible results. This work improved the reliability and predictability of optimization processes within the ITK codebase.