Developed a command-based configuration flow for the intelligent-machine-learning/dlrover repository, focusing on simplifying distributed training deployments. Leveraging Python and backend development skills, introduced the by_dlrover_run_cmd() method to generate DLJob configurations directly from command strings, supporting both dlrover-run and torchrun workflows. This approach enabled declarative, repeatable setups for Ray-backed training jobs, reducing deployment friction and streamlining onboarding for new users. The work emphasized robust testing to ensure reliability and maintainability. By shifting to a command-driven configuration model, the contribution laid a foundation for faster experimentation and easier scaling of distributed machine learning workloads within the dlrover ecosystem.