How gpu-mcp-server gives AI agents direct, real-time access to NVIDIA GPU telemetry for troubleshooting and autonomous operations.

AGENTS.md is a simple, open format for guiding coding agents. Think of it as a README for agents.

Design walkthrough for building a read-only multi-cloud Kubernetes inventory across OKE and AKS using a goose recipe, supervised shell access, and least-privilege credentials.

Learn how GPU-aware autoscaling helps Kubernetes respond faster to AI workloads, reduce idle GPU costs and protect inference performance.

Modern infrastructure has become excellent at provisioning compute, but artifact delivery is still a...

Getting Started with HAMi: GPU Virtualization on Kubernetes Running multiple AI workloads...

How gpu-mcp-server gives AI agents direct, real-time access to NVIDIA GPU telemetry for troubleshooting and autonomous operations.

The next FinOps challenge isn’t just tracking AI spend, but understanding which workflows create it, how agents amplify it, and whether the business value justifies the cost.

A deep dive into the dataplane — from Pod-to-Pod routing through veth pairs and Linux bridges, to Service load balancing.

How two open-source projects turn OKE GPU nodes into infrastructure that observes itself, scales...

How to deploy an MCP server on Oracle Kubernetes Engine that gives AI agents real-time GPU metrics -...

Learn about the five-dimensional design space in modern LLM serving, including tensor, pipeline, expert, data, and context parallelism

I had Prometheus + DCGM Exporter running on my OKE cluster. It gave me GPU utilization, memory usage,...

My GPU inference image was 8GB. Building it in GitHub Actions took 14 minutes on the free runner....

I was spending ~$3,300/month on three A10 GPU instances for a mix of staging inference, batch...

A colleague asked me to set up a private LLM endpoint their team could use for code review...

If you run GPU workloads on Kubernetes — vLLM, Triton, training jobs, or the newer agentic inference stacks — you’ve probably hit a familiar problem: the default autoscaling path still reasons about…

You can’t secure AI agents with traditional role-based access control. Instead, you need a zero-trust autonomous pipeline.

The models are ready. Your pipes aren’t.

Last month I needed to stand up a Llama 3 inference endpoint for an internal tool. The requirements...

The first time I deployed vLLM on OKE, the pod took 12 minutes to become ready. The model download...

Author: Pavan Madduri, Senior Cloud Platform Engineer & CNCF Golden Kubestronaut

Author: Pavan Madduri, Senior Cloud Platform Engineer & CNCF Golden Kubestronaut

Learn how KEDA enables GPU autoscaling for AI inference workloads on Kubernetes, reducing latency while cutting idle GPU costs.

Learn how KEDA enables GPU autoscaling for AI inference workloads on Kubernetes, reducing latency while cutting idle GPU costs.

The importance of formal verification in deploying Kubernetes manifests to ensure operational stability, and the limits of traditional tools.

The importance of formal verification in deploying Kubernetes manifests to ensure operational stability, and the limits of traditional tools.

If you run GPU workloads on Kubernetes — vLLM, Triton, training jobs, or the newer agentic inference stacks — you’ve probably hit a familiar problem: the default autoscaling path still reasons about…

Kubernetes v1.35 introduces in-place resource resizing, revolutionizing how stateful workloads are managed.

Discover how ArgoCD v3 transforms GitOps from simple state reconciliation to managing complex, multi-cloud environments.

Explore how the role of SRE is transforming with Agentic Ops, integrating technologies like eBPF, LLMs, and Kubernetes Operators.

Every GPU inference container has the same problem: Kubernetes HPA can’t see the GPU. You scale on CPU and memory while your GPU sits at…

Docker shipped two features in 2026 that fundamentally change how we run AI workloads locally: Docker Model Runner and Docker Sandboxes…

By Pavan Madduri

If you spend any time on Tech Twitter or LinkedIn, you are probably drowning in the phrase “AI Agents.” But if you strip away the marketing…

“The Boy Who Cried Wolf” is the oldest story about monitoring systems ever written. If the alarm goes off every five minutes for a minor…

Tags: #DevOps #Kubernetes #SRE #SoftwareEngineering

From Code to Production: A Complete Reference Architecture Using Only Open Source CNCF Tools Published on December 31, 2025 | Reading Time: 25 minutes 🚀 Introduction: The Cloud-Native Puzzle Throughout our 30-day CNCF project series, we've explored individual technologies—each solving specific challenges in cloud-native infrastructure. But here's the question every platform engineer asks: How do these pieces fit together? Today, we're building something special: a…

🛡️ Day 30: Kyverno - Kubernetes Policy Engine That Speaks YAML Validate, Mutate, Generate & Secure Your Clusters Without Learning a New Language Published on January 5, 2026 | CNCF Project Series - 🎉 GRAND FINALE! Connect with me on LinkedIn 🎊 Celebrating 30 Days of CNCF Projects! This marks the grand finale of our journey through the cloud-native ecosystem. Thank you for joining me on this incredible exploration of CNCF technologies!

🌐 Day 29: Kgateway - The Next-Generation Kubernetes API Gateway Feature-Rich Gateway API Implementation Powered by Envoy Proxy Published on January 4, 2026 | CNCF Project Series Connect with me on LinkedIn 🎯 What is Kgateway? Your Kubernetes cluster needs a front door—an intelligent gateway that routes external traffic to the right services, enforces security policies, handles rate limiting, and transforms requests.

📊 Day 28: Grafana - Where Your Data Becomes Beautiful Insights Visualize, Alert, and Explore Your Metrics, Logs, and Traces Published on January 3, 2026 | CNCF Project Series Connect with me on LinkedIn 🎯 What is Grafana? Data is everywhere—metrics from Prometheus, logs from Elasticsearch, traces from Jaeger, business data from PostgreSQL. But raw data is meaningless without visualization. When an incident strikes at 3 AM, you need dashboards that instantly reveal what's happening.

🔥 Day 27: Prometheus - The Heartbeat of Cloud-Native Observability Monitor Everything, Alert on Anything, Scale Without Limits Published on January 2, 2026 | CNCF Project Series Connect with me on LinkedIn 🎯 What is Prometheus? In a world of microservices, containers, and dynamic infrastructure, knowing what's happening inside your systems isn't just helpful—it's survival. When your e-commerce platform slows during Black Friday, when your API latency spikes at 3 AM, when your Kubernetes pods mysteriously restart—you need answers instantly.

🚢 Day 26: Harbor - The Enterprise Container Registry You Can Trust Secure, Scan, Sign, and Scale Your Container Images Published on January 1, 2026 | CNCF Project Series Connect with me on LinkedIn 🎯 What is Harbor? Your container images are the lifeblood of your applications. They contain your code, dependencies, and secrets. But where do you store them securely?