Product judgment
I've owned roadmaps and MVP scope myself, so I know exactly how tempting it is to overbuild. A lot of my job is helping teams not do that.
About
For fifteen years I've shipped software. Now I work on how you trust what AI builds.
Originally from Mexico, now an Icelandic citizen based in Reykjavík. Most recently I co-founded dala.care, a workforce platform for home healthcare providers. There I spent three years on the visit scheduling core, the caregiver mobile app, and the occasional 2 a.m. debugging session.
These days I'm deep in the question of how teams ship with AI agents without giving up the ability to prove the work is right. I write about it and build the tooling for it in the open.
Technical depth
I still write code, mostly Go, React, and AWS, plus more agent tooling than is probably healthy. When I say something is feasible, it's because I could build it myself.
Execution focus
Strategy that doesn't survive contact with the sprint board is decoration. I stick around to make sure things actually ship.
Where the experience comes from
At dala.care I worked across product direction, engineering, scheduling and coordination systems, infrastructure, and the AI-assisted development setup that let a small team move like a bigger one.
I also build and maintain Beislið, Memento Vault, and Rondo, open tools for making AI-assisted engineering dependable day to day. They started as itches I needed to scratch and turned into the backbone of how I work.
I've given talks on agentic development at Kolibri, Gangverk, and Orkuveitan, all circling the same question: how teams can use AI without losing judgment, context, or control. I also showed an early version of my agentic setup at Skelin. Watch the video, or see the Kolibri talk's event page.
Earlier work includes iOS apps for prosthetic knees at Össur, security systems for Sotheby's, and audio streaming at CBS Radio. You learn a lot about risk when your code runs inside someone's knee.
How I work
- Start from what the customer actually does, then work backward to the technology. That includes the AI parts.
- Prefer small steps that produce evidence over grand roadmaps that never meet a user.
- Get product, design, and engineering into the same conversation instead of passing documents between them.
- Stay close to the code. Strategy drifts the moment it stops being checked against reality.
Off the clock
Board games. Heavy euros, thematic dungeon crawlers, anything with interesting decisions. Always looking for people to play with in Reykjavík.
Tabletop RPGs. Running and playing. The overlap of storytelling, improv, and system design keeps pulling me back.
Cooking. I treat the kitchen like a codebase: iterate fast, taste often, occasionally produce something genuinely great. Latin roots, Icelandic ingredients, whatever catches my eye.
When I'm not doing any of that, I'm probably walking Miles (long hair mini dachshund, very important team member).
Say hi
Reading the series or using the tools?
I like hearing how the writing and the tooling land in the real world, especially when you disagree with something.
Get in touch