Hey, I'm Sam. I build things, break things, figure out why they broke, and write it up so the next person doesn't have to.
A bit about me
I'm a Lead Cloud Architect at Rackspace Technology, based in Edmonton, Alberta. I lead a team of seven engineers and spend most of my time designing AWS environments that actually hold up — secure, cost-efficient, and built to last. I've been doing this kind of work for about 15 years, though it didn't start in the cloud.
I've been a sysadmin at a forestry company managing 4,500 users across 150 North American sites. I've launched weather balloons in the Arctic at Environment Canada (seriously — Alert, Nunavut, one of the most remote places on Earth). I ran two restaurant franchises for a decade. I did L2 support at IBM before most of the tools we use today existed. Each of those things taught me something about how systems actually work under pressure, and I carry all of that into the cloud architecture work I do now.
On the certification side: AWS Solutions Architect Professional and Security Specialty. I also hold a certificate in Information Systems Management from Athabasca University and studied Electronics Engineering at NAIT.
What this site is
gitrdun.net is where I document real work — not polished tutorials, not vendor marketing. Actual problems from actual AWS environments, with the wrong turns included. The name comes from the thing I keep telling myself when something is hard and underdocumented: just get it done.
A lot of the content here exists because I spent hours or days tracking down something that should have taken twenty minutes, found one buried forum post from 2019, and thought: someone should write a proper explanation of this. That someone is me now, apparently.
How I use AI in this work
I use Claude Code as a genuine engineering partner — not to generate boilerplate, but to work through complex problems faster than I could alone. Multi-agent pipelines for research and validation. Structured data passing between agents. Python for deterministic rendering, not LLMs guessing at formatting.
The rule I've landed on: AI handles reasoning at scale, scripts handle anything that needs to be exact. Everything I publish has been verified in a real environment. If it didn't work, it doesn't go up.
The thing I'm most proud of building lately
An end-to-end identity pipeline: Entra ID SCIM provisioning into AWS IAM Identity
Center, ABAC session tags flowing through to SSM Session Manager, running as the
authenticated AD user on a domain-joined Linux instance. Real users, real domain,
proven with whoami = alice@gitrdun.net. Built in a day. Documented
the same week. It's the kind of thing that used to take a team and a sprint.
It doesn't replace engineering judgment. It removes the ceiling on how much ground one engineer can cover in a day. The judgment still has to be yours — and so does the accountability when something goes wrong.