Built a telemetry platform that collects GPU/CPU metrics (NVML), performs anomaly detection, and surfaces ML-based failure predictions. Integrated AWS S3 for time-partitioned metric storage and CloudWatch for custom monitoring metrics and alerting.