Engineering notes from production. AI agents, platform infrastructure, and the systems that compound.
// the work
I lead Site Reliability Engineers, Database Administrators, and Full Stack Developers, keeping production systems running while building the infrastructure that powers them. The technology and AI are the craft I love. The business they drive is what actually motivates the work.
Daily-driving open-source agent frameworks, running local LLM tooling with Ollama, and shipping upstream fixes when something breaks.
Cloudflare Workers in production with KV caching, debugging edge cases at scale, and optimizing for cost and reliability on real traffic.
CLI tools that remove friction, from VPN control to Pomodoro timers to AI-powered shell command generation. Small surface, durable leverage.
Frameworks for onboarding tech leaders, scaling engineering teams, and maintaining technical depth as an executive. Public, not theoretical.
Shipped code and led teams across Aerospace/Defense, SaaS, Retail, and Education. Four industries, four reliability and compliance regimes.
This blog is the lab notebook. I write about what I'm debugging, what I'm building, and what I'm learning, for engineers who want the actual answer.
// systems shipped
$4.2M
Savings delivered
99.95%
Production uptime
40%
Incident reduction
SRE 路 DBA 路 FSD
Teams led
// in flight
Open-source contribution and a small applied-AI venture. Both quiet, both real.
I daily-drive Hermes Agent on my own server and ship fixes upstream when something breaks. It's the agent framework I reach for, and the posts about it are the long-form version of those debug sessions.
Practical AI agents and automation for local service businesses. The parts that earn their keep on day one: scheduling, triage, follow-up, ops reporting.
// deep dives
Long-form notes on what I'm debugging, what I'm building, and what I'm learning. Written for engineers who want the actual answer.
Seven phrases that signal leadership maturity, ownership, and strategic thinking. The communication patterns that separate executors from strategists.
Hermes Agent kept timing out on GLM-5.1 mid-tool-call. The culprit wasn't Hermes - it was a 30-second idle timeout on api.z.ai colliding with how GLM batches tool-call output. Here's the diagnosis, the one-line fix, and the upstream PR.
A practical guide to getting started with Hermes Agent by Nous Research - installation, configuration, the skills system, memory, cron jobs, and the tips that actually make a difference.