👋 Hi! I'm Bibin Wilson. In each TL;DR edition, I share practical DevOps, Kubernetes, Cloud, and MLOps resources to help you stay up to date. Subscribe here to get every TL;DR edition in your inbox.
This Week in DevOpsCube ☕
Claude apps gateway for AWS
How Kubeflow Trainer Transforms AI Model Training
Is a Pod the Right Deployment Unit for AI Agents?
How Does Amazon EKS Scale Kubernetes?
How did Discord simplify multi-GPU training for ML Engineers?
Building a Multi-Cluster Kubernetes Platform with GitOps
DevOps Remote Jobs
and more..
☕ Grab a coffee and catch up on this week’s DevOps, MLOps, and AI insights and resources.
📢 Latest in DevOps
Kubernetes Introduced a new Plugin to manage Kubeflow - The new Headlamp plugin brings Kubeflow Pipelines, Katib, Trainer, and Notebook management into a single Kubernetes interface.
AWS Launches Claude Apps Gateway - AWS introduced Claude Apps Gateway, a self-hosted control plane that centralizes authentication, access policies, usage tracking, and cost controls for Claude Code and Claude Desktop running on Amazon Bedrock.
GitHub Launches the Copilot App - Now you can manage AI coding agents, parallel development tasks, pull requests, and software delivery workflows from a single interface.
🎯 Deep Dive: How Kubeflow Trainer Transforms AI Model Training
In this detailed edition, I have covered,
A look into distributed training
Kubeflow Trainer architecture (TrainJob, Runtimes, JobSet)
Hands-on: Install Trainer and run your first distributed TrainJob
How to run ML Training Jobs on GPU Nodes
Which Models Actually Need GPUs
By the end, you will know exactly how enterprise teams run distributed training on Kubernetes and what your job is as the MLOps engineer behind it.
👉 Read it Here: Kubeflow Trainer Transforms AI Model Management
📖 Worth Reading This Week
How Amazon EKS Scales Kubernetes - Learn how Amazon EKS scales Kubernetes by optimizing etcd, improving control plane performance, and handling large clusters more efficiently.
CI/CD Supply Chain Threat - Possible ways attackers can attack insecure CI/CD workflows and security practices that help prevent the attacks.
Spec-Driven Development is Reshaping Software Development - An overview of how Spec-Driven Development shifts software engineering from writing code to using AI agents to design, build, and maintain applications consistently.
Building a Multi-Cluster Kubernetes Platform with GitOps - A real-world case study on building a multi-cluster Kubernetes platform using Argo CD, with automated cluster onboarding and scalable cluster management.
Miasma Supply Chain Attack Targets AI Coding Agents - A breakdown of the Miasma supply chain attack that compromised Microsoft GitHub repositories by targeting AI coding agents through malicious repository configuration files.
⚙️ Tip: ImageVolume in Kubernetes
With Kubernetes ImageVolume, you can treat data like container images.
The ImageVolume feature lets you mount the contents of OCI images as Kubernetes pod volumes.
Here is a practical use case.
When deploying an AI model, it is usually retrieved from sources such as cloud object storage or external URIs. Packaging model data into OCI images makes it much easier to manage and switch between models.
Also, with subPath support, you can have multiple models within a single imageVolume. I first spoke about this feature when it was in alpha. Now, it is a stable feature.
I have published a hands-on guide to help you learn this concept practically. It covers the following:
What is ImageVolume
Building an ImageVolume with a sample model
Deploying an app to access the model via ImageVolume
Using subPath with ImageVolume
I have prebuilt images to make it easy to try it out.
👉 Read it Here: ImageVolume in Kubernetes
🤖 AI Infrastructure
Is a Pod the Right Deployment Unit for AI Agents? - A discussion on whether Kubernetes Pods are the right execution model for AI agents, and explores alternative runtime abstractions for managing agent lifecycles.
Why Kubernetes Is Becoming the Runtime for AI Agents - A closer look at how Kubernetes supports AI agent deployment through modern runtime and platform architecture.
How Discord Simplified Multi-GPU Training for ML Engineers - A case study on how Discord built a Kubernetes-based distributed ML platform using Ray, KubeRay, and Dagster to simplify multi-GPU training and orchestration.
Isolating AI Workloads in Multi-Tenant GPU Clusters - Learn how to isolate AI workloads in shared GPU Kubernetes clusters using Kubernetes-native isolation, NVIDIA MIG, and virtual clusters for secure multi-tenancy.
Securing AI Agents with NGINX and OpenTelemetry - Learn how to secure AI agents by controlling network traffic, monitoring agent activity, and improving observability using NGINX and OpenTelemetry.
🤖 Kubernetes Job vs. JobSet
A Job/Cronjob runs one specific batch workload. For example, ETL, backups, report generation, etc.
A JobSet, however, runs multiple coordinated Jobs together as a single distributed workload. Meaning each job in the set can have its own pod template and related configurations, but they behave as a single unit.
Behind the scenes, JobSet runs its child jobs in indexed mode. Meaning every pod gets a stable index and hostname, such as worker-0, worker-1, etc.
Also, JobSet includes a headless service so that pods can discover one another. The key use cases for JobSet are AI/ML and HPC workloads. For example, in Kubeflow trainer, Jobset is used to run distributed training workloads.
🛠️ DevOps Tool of the Week (KubeArmor)
KuberArmor is an open-source security tool that controls what can run, what files can be accessed, and what network calls can happen inside your workloads at the kernel level.
Instead of just alerting, it blocks the action before it runs.
Here is what it does👇
It blocks unexpected process execution or file access inside a pod before it happens.
Enforces least-permissive access with process and network whitelisting per workload.
Locks down sensitive paths like cert bundles and secrets so they can't be read or changed.
Applies hardening policies mapped to MITRE, CIS, and STIG without writing them yourself.
Provides detailed runtime visibility using eBPF with pod, container, and namespace information.
👉 GitHub Repo: KubeArmor
🎓 Complete Kubernetes & CKA Course
10,000+ engineers have learned through DevOpsCube courses.
From container fundamentals to CKA preparation, every course is self-paced and grounded in real-world 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.
👉 Coupon: Use code COMMUNITY50 at checkout.
Note: This is a community-only offer! Make use of it!
💼 DevOps Remote Jobs
ZettaMine Labs - AI DevOps (5+ Yrs)
Five Data Products & Solutions - DevOps AI Cloud Engineer (4+ Yrs)
Eurofins - DevOps Engineer
OVI - AIOps/MLOps Engineer (4+ Yrs)
EXL - MLOps Engineer (2-4 Yrs)
Quillbot - MLOps Engineer II (3+ Yrs)
XCaliber Health - Platform Engineer (2+ yrs)
Hire Feed - AWS Engineer - Cloud
Scoutit - AWS Cloud Engineer
Avahi - DevOps Engineer
Subscribe to Learn DevOps & MLOps
20,000+ trusted Engineers read our newsletter, where we simplify DevOps, MLOps and Kubernetes into practical, easy-to-follow guides.
Or subscribe hereDevOpsCube team:
Lead Editor: Bibin Wilson / Designer: Aljin / Tech Researcher: Aswin
Interested in advertising with us? Get in touch
Need DevOps consulting? Contact us
Want to learn from courses? Check our courses
Forwarded this email? Join our mailing list
