Blog Posts 13
Technical articles on the official AWS blogs and personal writing on Substack
Not all "AI Chips" do the same thing
There is no universally best AI accelerator, only one whose trade-offs match your workload. A walk through eight chip families against the three walls of compute, memory capacity, and memory bandwidth: GPUs for generality, Groq's LPU for decode latency, TPUs for dense matrix ops at scale, NPUs for the edge, and DPUs for infrastructure offload.
Why I Stopped Thinking About AI Infrastructure as "Just GPUs"
AI infrastructure has evolved past stacking GPUs. As workloads shift from chatbots to agents, success means balancing three coordinated pools — accelerators for tensor math, CPUs for orchestration logic, and DPUs for networking — tuned to whether you're training, prefilling, or decoding.
Deloitte optimizes EKS environment provisioning and achieves 89% faster testing environments using Amazon EKS and vCluster
Case study on how Deloitte combined Amazon EKS with vCluster to cut test environment provisioning from 45 minutes to under 5, freeing ~500 QA hours per year and consolidating infrastructure costs.
Navigating enterprise networking challenges with Amazon EKS Auto Mode
How EKS Auto Mode tackles enterprise Kubernetes networking — VPC CNI optimization, pod density, eBPF-based network policies, and hybrid connectivity at scale.
Navigating the NGINX Ingress retirement: A practical guide to migration on AWS
A practical migration guide for teams impacted by the Ingress NGINX retirement, covering planning and execution on Amazon EKS and hybrid environments.
Multi-VPC Load Balancing for Amazon EKS workloads
Practical walkthrough to achieve enhanced security isolation, network segmentation, and more flexible network design.
Enhancing Developer Productivity: Finch's Support for Dev Containers and the Finch Daemon
Guide to Finch's new dev container support and the Finch Daemon, streamlining container build and run workflows.
Announcing Finch on Linux for Container Development
Finch's Linux support announcement — build, run, and publish containers on Linux with familiar Finch commands.
Unlocking Benefits with Bottlerocket: A Purpose-Built Container OS
Walkthrough of Bottlerocket's fleet management capabilities and the key pillars supporting its development plan.
Cordial's journey implementing Bottlerocket and Karpenter in Amazon EKS
Case study of Cordial's implementation of Bottlerocket and Karpenter in Amazon EKS to streamline operations.
Proactive scaling of Amazon ECS services using ECS Service Connect Metrics
Leveraging Amazon ECS Service Connect metrics for proactive scaling of ECS services.
Multi-account infrastructure provisioning with AWS Control Tower and AWS Proton
Using AWS Control Tower and AWS Proton for multi-account infrastructure provisioning with centralized control.
Security features of Bottlerocket, an open source Linux-based operating system
Exploration of Bottlerocket's security features and built-in controls for secure container environments.
How to automate Puppet manifest testing and delivery in AWS OpsWorks
Setting up CI/CD for Puppet manifests using CloudFormation, CodeCommit, CodeBuild, CodePipeline, and Systems Manager.