In March 2025, Paul Cook worked on the InsightSoftwareConsortium/ITK repository, focusing on improving the reproducibility of optimization workflows. He addressed non-determinism in multi-start optimization by resetting the learning rate for each run and introduced an InitialLearningRate parameter, along with mechanisms to store and restore the optimizer’s state at the start of each execution. Using C++ and applying principles from optimization algorithms and software engineering, Paul also developed a unit test to verify that repeated optimizations with identical starting parameters yield consistent results. This work enhanced the reliability and predictability of ITK’s optimization routines, demonstrating careful attention to workflow integrity.