This is my working environment for learning AI infrastructure hands-on. Kubernetes clusters, GPUs, networking, storage, observability, inference, automation, and the occasional broken configuration. I record the experiments, the problems, the fixes, and the measurements.
Current Work
GPU Scheduling — Kubernetes GPU resources, device plugins, node labels, taints, affinity, priorities and preemption. Status: In progress
NVIDIA GPU Operator — Installing and understanding the components behind the NVIDIA GPU stack on Kubernetes. Status: In progress
GPU Observability — DCGM, Prometheus, Grafana and infrastructure-level GPU telemetry. Status: Building
Inference — Exploring the infrastructure requirements for serving AI models efficiently. Status: Exploring
Performance — Networking, storage, CPU/GPU utilization, latency and throughput. Status: Ongoing
The lab is not a demo environment. It’s where I try things, break things, measure them, and write down what I learn.
- DCGM GPU Telemetry in Practice
October 10, 2026
Installing DCGM 4.7, verifying GPU health, and watching live SM and memory counters under a sustained PyTorch workload.