This Week in DevOpsCube ☕

☕ Grab a coffee and catch up on this week’s DevOps, MLOps and AI insights and resources.

If you want to level up in MLOps, Feature Store is a must-know concept. It is one of the core building blocks of MLOps.

In our latest MLOPs newsletter edition, we covered the following.

- What is a feature store actually
- Offline vs Online features (very important)
- What is feature registry.
- Feast feature store architecture
- Role of DevOps engineers in feature management.
- Hands-on Feast Feature Store setup on Kubernetes

👉 Read It Here: Feature Store Explained for DevOps Engineers

The best way to understand a ML feature store is by setting it up and manage features yourself. We have a detailed hands-on guide that covers the open-source Feature Store called Feast.

In this guide, you will learn:

- What is Feast?
- Key Feast components
- Feast Operator setup on Kubernetes
- Configure offline and online stores.
- Use a simple Python script to verify feature serving
- How to measure feature serving latency metrics like p50, p95, and p99

👉 Read It Here: Feature Store Setup on Kubernetes

You won't see this 65% Discount again until Cyber Monday👇

Get flat 50% off on individual, certifications like CKA, CKAD, CKS using the following coupon.

Use code MM26BUNCT to save up to 60% on the following Kubernetes certification bundles.

⏳ Once this offer expires, it’s gone. Grab it while you can.

When deploying a Kubeadm based kubernetes cluster on AWS with Calico CNI, you may encounter a connection timed out issue between Pods and CoreDNS.

We encountered this issue and we have created a detailed blog that explains,

👉 Read It Here: Troubleshooting Calico CNI on AWS

10,000+ engineers have learned through DevOpsCube courses.

From container fundamentals to CKA preparation, every course is self-paced and built around real scenarios.

This is not a long video lecture series.

The CKA course is text-based, illustration-rich, and designed for faster learning and quick revision whenever you need it.

👉 Use code FLASH40 to get 40% OFF today.

Image optimization is not just an infrastructure problem. It is a collaborative effort between data scientists, developers, and DevOps engineers.

Because each role owns a different layer of the bloat.

For example, a DevOps engineer cannot safely remove a library without confirming with the data scientist whether the model actually needs it. Real optimization happens when all teams sit together and ask: What does this image actually need at runtime?

In our case, an image used in a Kubeflow pipeline project was 3.17 GB.

In this blog, you will learn how the image was optimized to 354 MB (an 89% reduction) and the reasoning behind every change.

👉 Read it Here: Kubeflow Docker Image Optimization

If you are working with Helm charts, you might have seen a helpers file inside the /templates folder. Most people ignore it without understanding what it is used for while deploying a chart.

In our guide, we have covered,

👉 Read it Here: Reusable Helm Templates Using _helpers

This video explores how Uber uses the Model Context Protocol (MCP) to scale AI agent workflows. For a DevOps engineer, the key takeaways involve:

👉 Watch the Video Here: How Uber Runs 60,000 AI Agent Tasks

If you are a DevOps engineer trying to understand how AI fits into infrastructure, automation, and developer workflows, these Claude courses are worth checking out.

Anthropic has structured learning paths around:

👉 Start Here: Free Claude Courses

Running Kafka in production is great until you need to actually look inside it.

Kafbat UI is a free, open-source web UI to monitor and manage Apache Kafka clusters.

👉 Start Here: Kafbat UI