Platform engineering notes.
Practical writing on running AI workloads in production: GPU platforms, RAG infrastructure, edge fleets, and the operational patterns that keep systems reliable.
Published posts
Canary Is a Cohort, Not a Percentage
Five percent of twelve Jetsons is not a canary. Edge rollouts need named groups of devices, health gates between them, and a blast radius you can actually reverse.
Read postEdge Failure Modes Are Network Failure Modes First
If you cannot tell partition from crash, you do not have an edge platform. Inference often keeps running while GitOps, scrapes, and rollouts stall on a bad link.
Read postWhat Platform Engineering Actually Means for AI Teams
The gap between a working model and a reliable production system is a platform problem. Here is what that layer covers, what it does not, and when your team needs it.
Read postThe Production Readiness Gap in AI Platform Engineering
A working model is not a production system. Engineering leaders need a clear view of what is actually in place before AI workloads face real traffic, real cost, and real incident pressure.
Read postFrom Cloud GPU Cluster to Edge Fleet: One Operational Model
How GitOps, observability, and controlled rollouts apply whether workloads run in a datacenter or on NVIDIA Jetson in the field.
Read postReady to take your AI workloads to production?
Let's talk about your platform: cloud, hybrid, or edge. Start with a short, no-pressure conversation.