L Singh on Muck Rack

L Singh

Sydney
As seen in: Medium, Wiley Online Library, Hindustan Times, Flipboard, Royal Society of Chemistry, Tone Deaf, The BRAG, Economic and Political Weekly, ACI Materials Journal, tmrw and
Is this you? As a journalist, you can create a free Muck Rack account to customize your profile, list your contact preferences, and upload a portfolio of your best work. Claim your profile

Get in touch with L

Contact L, search articles and posts on X, monitor coverage, and track replies from one place.

Learn more about Muck Rack

Actions

Share this page

Is this you?

As a journalist, you can create a free Muck Rack account to customize your profile, list your contact preferences, and upload a portfolio of your best work.
Claim your profile

Articles

Allocated but Idle: The Network and Storage Stack Behind Every AI Workload on Kubernetes

|
By L Singh
| DevOps Blog
Press enter or click to view image in full size Your pod is Running. Your GPU is allocated. DCGM still says it is idle most of the time. If you followed the first post and got the NVIDIA stack installed, your training pod will now schedule. The driver works. The container toolkit works. The device plugin advertises nvidia.com/gpu. The scheduler finds a node. The pod requests its GPU and gets it. Everything in the NVIDIA Kubernetes stack lights up green.
Open in Who Shared Issue with article?

Green Dashboards: Production Monitoring and Logging for GPU Workloads on Kubernetes

|
By L Singh
| DevOps Blog
Press enter or click to view image in full size Kubernetes GPU workload Monitoring Dashboards Something just broke. Your dashboards are green. The customers know first. If you followed the first three posts, your cluster can schedule GPUs (part one), feed them properly with network and storage (part two), and grow and shrink to match demand (part three). The infrastructure is in place. Workloads run. The cluster scales itself. And one day inference p99 latency goes from 200ms to 8 seconds.
Open in Who Shared Issue with article?

From Zero to 256 GPUs: Deploying and Autoscaling AI Workloads on Kubernetes

|
By L Singh
| DevOps Blog
Press enter or click to view image in full size Deploying AI workloads to Kubernetes Your code is in Git. Your cluster has zero GPU nodes. Your training job is waiting. If you followed the first two posts (post1 and post2), you have a Kubernetes cluster that can schedule GPUs and feed them properly. The whole stack is sitting there, ready. But nobody has actually run anything on it yet, and you have not decided how they will. That is part three.
Open in Who Shared Issue with article?
See all 570 licensed articles (573 Total)
 •  Privacy  •  Terms  •  Security and Legal  •   •   •   • 

© 2026 Muck Rack

SOC 2® Certified