Developed and integrated production-grade Prometheus observability metrics for the NVIDIA/TensorRT-LLM repository, focusing on comprehensive monitoring of iteration statistics, configuration details, token counters, and phase histograms. Leveraging Python and backend development expertise, the implementation provided high-fidelity telemetry with minimal overhead, seamlessly fitting into the existing metrics collection pipeline. This work enabled end-to-end monitoring, supporting SLA reporting, faster troubleshooting, and proactive performance tuning for large language model deployments. By delivering a single, verified commit, the developer enhanced the system’s capacity planning and reduced mean time to resolution, ensuring robust and actionable insights for ongoing operational excellence in LLM infrastructure.