From kubectl to Gradient Descent: Why I'm Betting My Career on AI

I’ve spent a decade making infrastructure disappear — pipelines, IaC, container platforms, the invisible plumbing that lets other people ship. It’s been a good decade. I’m also, deliberately, walking toward something different: AI is becoming the center of my work, and next year I start a Master’s in Artificial Intelligence at UNC Charlotte.

This is the “why now.”

The pivot started at work, not in a classroom

The classroom is next year. The turn already happened, on the job.

At Vanguard I became my team’s AI SME — not because I trained a model from scratch, but because I got obsessed with a more immediate problem: AI coding assistants are only as good as their context. Drop a general-purpose model into a sprawling enterprise codebase and it writes plausible code that doesn’t fit your systems, your standards, or your infrastructure.

So I built the missing half. I created internal context repositories — curated knowledge that gives an assistant like Claude a real understanding of how our infrastructure and applications actually work — so the code it generates fits the way we build. That’s the work I find genuinely exciting right now: not treating AI as a magic box, but engineering the scaffolding that makes it useful on real, messy, production systems.

Turns out a decade of DevOps is unreasonably good preparation for this. Knowing how systems actually fit together is exactly what lets you feed an AI the right context.

Why go get the degree

If the work is already happening, why the Master’s?

Because I can feel the edges of what I know. I can wire up retrieval, engineer context, and ship AI-assisted tooling that saves real hours. What I can’t yet do from first principles is the layer underneath — the math and theory of how these models actually learn. I’ve spent my life refusing to treat the systems I work on as black boxes. I opened up radios, servers, and hypervisors. I’m not about to make an exception for the most important technology of my career.

The degree is me doing what I always do: refusing to operate one level of abstraction above where I’m comfortable.

What this blog becomes

Expect more AI here — the honest kind. Not “AI will change everything” thought leadership, but real projects:

• Self-hosted models on hardware I’ve unlocked myself (my home AI server grew a sequel ).
• Context engineering and retrieval that actually improve generated code.
• Whatever the Master’s throws at me, worked through in public.

A decade in, I still get the same feeling from a model that finally behaves that I got from a pipeline that finally went green, or a homebrew circuit that finally transmitted. It’s all the same itch: understand the system, then make it do something real.

More soon.