Worked on the open-mpi/ompi repository to address a race condition in the MPI_Alltoall collective operation, focusing on high-concurrency parallel computing environments. Using C and leveraging expertise in low-level programming and MPI, introduced a synchronization barrier within coll_han_alltoall.c to prevent the send buffer from being overwritten while neighboring processes read from it. This targeted bug fix, prompted by a report from the Barcelona Supercomputing Center, improved the correctness and stability of message passing in distributed systems. Collaborated closely with external researchers to reproduce and validate the issue, ensuring the solution enhanced reliability for demanding parallel workloads without introducing regressions.