DueWeekly Tech: 01-09-2026 to 06-09-2026

Quote

It has been quite a while since the last Dueweekly post. Recently, I came across a few fascinating articles that I would love to share with you. If you would like to explore further, please check out the ones below.

AIOps / MLOps / LLMOps

1. Medium - The Four-Layer Stack for Running Agentic AI on Kubernetes 🌟 (Recommended)

• This great article helps guide you through the essential steps of running Agentic AI on a Kubernetes cluster. However, while it provides valuable information on frameworks, tools, and mindmaps to help you understand why Kubernetes is chosen for AI agents, it lacks hands-on guidance.
• In the first part of the session, the author explores the key factors to consider when deploying Kubernetes for your workloads, breaking down AI workloads into manageable layers—such as Inference, Training, and Agentic Systems. They also address the current limitations and gaps of using Kubernetes for AI, highlighting its added complexity and areas where it falls short.
• You will also discover several innovative tools, such as airunway (an AI Platform for Inference), kaito (a Kubernetes Operator for inference, fine-tuning, and RAG), alongside familiar solutions like Ray. Ultimately, the article offers a wealth of resources if you want to dive deeper into Agentic AI and orchestrate it on Kubernetes using AKS (Azure Kubernetes Service).

2. Blog - NVIDIA Jetson + K3s: Edge AI Cluster Tutorial (2026)

• A comprehensive tutorial designed to guide you closer to running AI inference on an edge cluster using K3s, covering rolling updates, health checks, GPU scheduling, and more. These articles outline numerous steps to seamlessly integrate them for easier implementation on a Jetson cluster, while explaining why Kubernetes was chosen.

Architecture

1. Blog - AI Factory: Bản chất, bức tranh toàn cảnh và thực trạng Việt Nam 🌟 (Recommended)

• The landscape of Nvidia’s AI Factory spans from major vendors to local players, with a specific focus on the Vietnamese ecosystem (featuring FPT, GreenNode, Viettel, CMC, and others). This article explores their current market positions and core strategic focuses, such as Serverless GPU, AI Notebooks, and K8s-native solutions utilizing VKS (VNG GreenNode) for high-performance inference, alongside low-latency, high-bandwidth distributed AI model training.
• Furthermore, the article examines the global GPU market layers, detailing which vendors dominate specific segments and how their business models influence the broader AI era—specifically the transformation of energy into computational tokens. It also highlights how the Platform Layer empowers diverse providers to leverage these opportunities and deliver advanced capabilities to their users.

2. Blog - Lạm bàn về… hướng đi nào cho AI Factory nội địa 🌟 (Recommended)

• Continuing the discussion on the AI Platform Landscape in Vietnam, this article outlines the challenges and the roadmap for local AI factories aiming to expand globally. However, it will not be an easy journey; the ultimate question is who will fund it, regardless of whether you develop your platform using MaaS, PaaS, IaaS, or SaaS models.
• Moving to the next question: what are the differences between local and global solutions? Operating on the same hardware, efficiency is the primary factor that makes this approach viable through GPU sharing, automatic reclamation or release of idle GPU workloads, and workload prioritization to ensure optimal inference and demonstrate full resource utilization. Beyond that, considerations must include delivery time, data governance for security, and containerization as a best practice for streamlined deployment, even for legacy systems.
• The optimal solution is to build a platform that integrates these methodologies, transforming the AI Factory into a comprehensive ecosystem. Here, users can find compelling reasons to handle deployment, cost management, observability, and governance, while maximizing the inference efficiency of AI models. Ultimately, this is not a race to the bottom on price; rather, it focuses on providing a robust platform and ecosystem that convinces enterprises to run their businesses on it.

Kubernetes

1. DevZero - Best Kubernetes VPA Alternatives for Production 2026

• The curated list of commerce solutions for scaling Kubernetes nodes explores the trade-offs and strengths of each platform. While it may not provide absolute clarity on every detail, it covers sufficient information regarding why VPA should be considered and how these solutions align with your Kubernetes cluster.

2. Cilium - Multi-Cluster Kubernetes Explained 🌟 (Recommended)

• A brand-new article on Cilium and Multi-Cluster in 2026 explores the what, why, and when of adopting a multi-cluster architecture. It also delves into various deployment patterns, such as Active-Active, Active-Standby, Cluster Per Environment, Cluster Per Region, and Hub and Spoke Topologies, to help you determine which approach best aligns with your requirements.
• Introduces the multi-cluster networking concept utilizing Cilium Clustermesh for seamless inter-cluster connectivity, cross-cluster pod-to-pod communication via Encapsulation Mode (Tunneling) or Native Routing (Direct Routing), global services with cross-cluster load balancing, cross-cluster network policies, and comprehensive multi-cluster observability powered by Hubble.

3. RedHat - Multi-cluster GitOps with the Argo CD Agent Technology Preview

• Introducing a novel method for running GitOps across multi-cluster environments using the ArgoCD Agent. This powerful enhancement secures multi-cluster ArgoCD operations, bridges environment and multi-cloud gaps, and resolves location-based challenges.
• This innovative approach maximizes resource efficiency, centralizes the entire GitOps model, and delivers exceptional visibility and insights for complex multi-cluster architectures.

4. CloudKeeper - Multi-Cluster GitOps with ArgoCD for Kubernetes

• A comprehensive guide on setting up a multi-cluster architecture using ArgoCD in a Hub and Spoke topology. It is detailed and easy to follow, allowing you to easily test it yourself if you are considering this stack adoption.

5. Medium - Deploy OpenStack on Kubernetes in under 60 minutes 🌟 (Recommended)

• This comprehensive guide walks you through building an OpenStack cloud on a Kubernetes cluster using K3s, Ceph, Proxmox, and OpenStack-Helm. This article details the entire journey from provisioning virtual machines from scratch to bootstrapping the Kubernetes cluster and deploying OpenStack, making it an invaluable read that will save you time.
• The guide leverages several interesting tools and mechanisms. For networking, OPNSense handles essential services such as DNS, load balancing, and virtual IP management, for Kubernetes managment, you can see him used Kite as modern dashboard. To help you seamlessly master and self-host these complex tools, the author also introduces GitHub - osh.

6. Medium - End-to-End DevSecOps on EKS: Every Stage Blocks or It’s Theater

• Operating a DevSecOps Pipeline on Kubernetes (EKS): The author outlines a clear conceptual contract using three core roles to keep the DevSecOps workflow running as an automated, interconnected chain: Gate (Trivy, Gitleak, Pre-commit, Cosign, Kyverno Admission) , Pager (Trivy), Memory (Detect Dojo).

Platform Engineering

1. Medium - Top 10 Platform Engineering Jobs & Skills Required

• The articles highlight several key areas to focus on, learn, and practice in order to shape your Platform Engineering career. To me, this serves as a great starting point for anyone looking to discover more keywords and delve deeper into the field.
• Additionally, you will find a couple of common models, architectures, and hot topics that platform engineers have already built and shipped, for example, AI Infrastructure (MIG, vLLM, SGLang, etc.), AI Observability (MLflow, Langfuse, etc.), etc. You can view this as a landscape to explore, but your personal path should be defined in greater detail. Pick an area you are passionate about and master it.