Zeke Lares
In June 2022 I walked into a company of a dozen people that had just raised its Series A. Butlr had a beautiful idea: understand how people use space through body heat, never through cameras. It also had what every young company has underneath the idea: a platform assembled at startup speed, held together by heroics, with a fleet about to grow faster than anything built to carry it.
One of my earliest commits, the seed of what became our shared engineering foundation, was literally titled Init HelloWorld. There is something honest about that. Nobody hands you an architecture. You start with hello world and you earn the rest.
The cleanup years
The early work was unglamorous on purpose. Authentication became one library instead of five opinions. Data access became one pattern. The schema got owned instead of inherited. Every service got the same deployment shape, the same metrics, the same alarms. None of this demos well, and all of it is the reason the platform could later grow by orders of magnitude without falling over.
Then came the migrations, and migrations became my craft. We moved the core datastore. We overhauled the device messaging layer, hardened it, and years later retired it entirely. We brought our whole time series history in house. Every one of those ran against a live fleet, for customers watching dashboards that could not be allowed to blink.
The best migration is the one your customers never find out happened.
Those were the sleepless years. I have debugged a gateway that went quiet after a cellular hiccup and never told anyone. I have chased data that arrived rotated, duplicated, or timestamped from the wrong decade. Debugging a fleet you cannot touch, spread across buildings you will never visit, teaches a specific humility: the field does not care how clean your code looked in review.
Thousands of sensors, and the things that stayed
The company grew to many times the size it was when I walked in, and the fleet grew from those first hundred or so sensors into thousands, streaming from buildings around the world. But the part I am proudest of is not the growth curve. It is that the earliest systems are still on duty. The reporting engine I wrote in my first months still answers customer queries. The alerting rail that watches over senior living residents still pages the moment something matters. The foundation that started as hello world now sits under every backend service we run.
Somewhere along the way I stopped counting what I added and started counting what I removed. Whole messaging layers, deleted. Dozens of tiny single purpose services, collapsed into a few efficient ones. Legacy remote access agents, uninstalled from the entire fleet. A codebase, like a company, gets healthier when it loses weight.
Create the repository. Build the first version. Run the migration. Delete the legacy. Stay on call for what you shipped. It is not a glamorous loop, but run it for four years straight and you end up with a platform that survives its own growth.
From sleepless nights to AI in the loop
The first half of my Butlr years happened before AI entered the engineering loop. Every migration plan, every incident review, every 3am log dive was purely human, and the hours showed it. The second half looks different. We put AI into the product itself, and just as importantly into the engineering loop: agents that help write and review code, replay harnesses that tune sensing for each individual device, models that watch telemetry streams no human could keep up with. I wrote about the discipline that makes this safe in Putting AI in the Critical Path. The short version: AI multiplied us, but only because the boring reliability work came first.
What the commit log taught me
Fifteen years into this career, my strongest belief is that a commit history is the most honest resume there is. Titles compress. Diffs do not. Mine says I bootstrapped the repositories, ran the migrations, deleted the legacies, answered the pages, and stayed long enough to watch the systems I started keep serving customers years later. That last part is the whole point. Anyone can ship. The craft is shipping things that keep working after you stop watching them.
Butlr grew from a dozen people into a company many times that size, and from its first sensors to thousands around the world. I grew with it, from an engineer who could build systems into one who could keep an entire company's systems standing while rebuilding them in flight. The sleepless nights were the tuition. The fleet still streaming this morning is the degree.