k0s Joins NVIDIA AI Cluster Runtime with a Validated H200 Training Recipe
An upstream k0s recipe brings signed H200 validation evidence to NVIDIA AICR. Learn what was tested, its Preview limits, and how to use or contribute a recipe.
Satyam Bhardwaj
Posts
Deploy and Validate GPU Stacks With NVIDIA AICR Data Packs
Package GPU stack configuration as an AICR data pack, deploy it through Helm, and retain verifiable validation evidence. A practical k0s/H200 walkthrough.
Satyam Bhardwaj
Modern BareMetal Provisioning When Redfish Implementations Disagree
A design for a Redfish proxy that handles vendor API differences through scripts.
Serhii Ivanov
Observability for DRA Drivers
How Prometheus metrics landed in kubernetes-sigs/dra-example-driver's instrumentation, client-go REST metrics and Helm-ready scraping on port 8080.
Vishal Anarase
Running AI Workloads in Hardware-Isolated Sandboxes with k0s and Kata Containers
Learn why standard container isolation fails for agentic AI workloads, and how k0s plus Kata Containers closes the gap with hardware-enforced sandboxing
Prashant Ramhit
Playgrounds of the Mind: Changing the Agentic Ops Trust Paradigm
Why the sandbox 'straightjacket' approach to securing AI ops agents is fundamentally flawed, and what a 'playground' paradigm—autonomy within bounded infrastructure, vetted workflows, oversight, and greylisted boundaries—looks like instead.
Randy Bias