During March 2025, Girasoley developed a pluggable collective backend framework for the facebookresearch/param repository, focusing on distributed training with PyTorch. Leveraging Python and the mixin pattern, Girasoley introduced a plugin-friendly architecture that allows users to integrate custom collective backends, including a new pytorch_ncclx_backend supporting dynamic all-to-all operations. The work included refactoring collective operation handling to ensure proper resource management and synchronization correctness, addressing challenges in distributed systems. By enabling user-defined kernels and simplifying backend integration, Girasoley’s contributions improved flexibility and performance for experimentation and collaboration, demonstrating depth in backend development and distributed systems engineering within the PyTorch ecosystem.