Worked on enhancing Slurm timeout handling within the NVIDIA/NeMo-Skills repository, focusing on backend development and configuration management using Python. Refactored the timeout logic to unify how durations are represented and processed across both Slurm and NeMo-RL workloads, improving reliability for batch job submissions. Developed utility functions to parse and format timeout values, ensuring compatibility between different components. Introduced a default timeout fallback in the cluster configuration, simplifying setup when partition-specific timeouts are not defined. The work emphasized maintainable code and robust configuration, drawing on DevOps practices to streamline cluster operations and reduce the risk of misconfigured job timeouts.