Worked on enhancing the NVIDIA/NeMo-Skills repository by implementing flexible container support within the run_cmd function. This feature allowed users to specify custom containers either through predefined cluster_config entries or by providing explicit container names, streamlining environment selection for both experimentation and production workflows. The approach focused on improving reproducibility and reducing setup friction, enabling more stable and repeatable runs. Leveraging backend development and DevOps skills, the work was carried out using Python and integrated seamlessly with existing deployment processes. The update addressed the need for customizable environments, supporting a more efficient and user-driven workflow for containerized experiments.