Worked on the isaac-sim/IsaacLab repository to enhance user-facing documentation, specifically targeting the challenges of multi-GPU training on Linux systems. Focused on troubleshooting NCCL errors, the work involved adding detailed notes and environment-variable workarounds to address CUDA memory access issues during distributed training. Leveraged reStructuredText to clearly communicate solutions, such as setting NCCL_SHM_DISABLE and NCCL_IB_DISABLE, ensuring users could achieve stable training without modifying runtime behavior. This documentation-only update strengthened the knowledge base for enterprise users, reduced support overhead, and improved the onboarding experience for those deploying distributed workloads. Key skills included documentation, multi-GPU training, and troubleshooting.