Shengqi focused on improving distributed training stability in the bytedance-iaas/sglang repository by addressing a critical bug in tensor parallelism rank management. Using Python and leveraging expertise in distributed systems, Shengqi modified the initialize_dp_attention function to retrieve local_rank directly from the tp_group object, rather than deriving it from tp_rank or tp_size. This targeted fix prevented misrouted attention across processes, reducing instability and debugging time in multi-node training environments. The work enhanced the reliability and maintainability of large-scale training jobs, demonstrating a deep understanding of distributed tensor parallelism and its operational challenges in real-world machine learning workflows.