Developed a dynamic Training Job API for the NVIDIA/NeMo-Run repository, enabling seamless support for both single-node and multi-node distributed training workflows. The implementation centralized job creation logic in Python, allowing the API to automatically select the correct endpoint and payload based on node count, which reduced complexity and improved maintainability. Comprehensive unit tests were added to validate both submission paths, enhancing reliability and test coverage. The work focused on backend development and API integration, laying a foundation for scalable distributed systems. This approach facilitated safer orchestration of training jobs and positioned the codebase for easier future enhancements and flexibility.