Notes

Writing archive

28 posts, 2012–2026: reliability engineering, platform architecture, migration constraints, and AI-assisted engineering.

Recent

Latest

Observability for Placement Decisions

A shared event contract makes placement decisions explainable across proxies, queues, workers, runtimes, and artifacts without centralizing sensitive workload content.

How I Write Postmortems

I write postmortems to explain how a failure reached production, what made the impact worse, and which changes will make the next incident easier to prevent or recover from.

Admission Control for AI Workloads

A target can be eligible and still not be ready. Admission needs a bounded capacity claim before competing workloads can start.

From the archive

Earlier writing, kept online for reference. The CentOS, Git deployment, and Linux migration pieces are dated but the approach still holds up; the personal essays from 2012–2014 don't reflect how I write or think now.