Worked on the carla-simulator/carla repository to streamline the Poisson disc sampling workflow by simplifying point generation and reducing metadata complexity. The approach involved removing random seed generation and eliminating per-point random metadata, which improved reproducibility and made the pipeline easier to maintain. By deleting the 'Density' metadata attribute and related code, the implementation reduced the overall metadata surface area, supporting more reliable and deterministic results. This refinement, developed using C++ and leveraging Unreal Engine, resulted in a cleaner, more maintainable codebase. The changes also facilitate faster onboarding for new contributors by making the algorithm implementation more transparent and easier to debug.