Infrastructure, Kubernetes, and applied AI
Notes on running AI workloads on infrastructure you control — Kubernetes, data platforms, and the operational details that decide whether a design survives contact with a cluster.
-
Five data engines, one namespace: a fraud pipeline on HPE Data Fabric
A medallion pipeline touching five data engines and three services, and the day I spent learning where the one-namespace abstraction has seams.
-
Deciding what crosses the link: an edge-to-core pattern on HPE Data Fabric
A field team on a satellite link cannot pull down everything headquarters has. That is not a bandwidth problem, it is a decision problem — and it is a good test of whether one platform really beats an assembled stack.
-
Building an agentic AI stack on Kubernetes Agent Sandbox: what it's for and what it's like to use
Prompt-based restrictions are a request, not a control. What it takes to move agent permissions into Kubernetes, and what running it actually teaches you.