I’ve been building things for over a decade. Production ML systems, distributed backends, recommendation engines serving tens of thousands of requests per minute. And for most of that time, I wasn’t writing any of it down.
That was a mistake.
Every time I solved a hard problem, it lived in my head — or worse, in some forgotten terminal session. Every time someone asked how I set something up, I’d have to reconstruct it from memory. Every new project, I’d rediscover something I’d already figured out a year earlier.
So I’m writing it down now. That’s Trailstack.
A Bit of Background
I’ve spent 10+ years in software and machine learning engineering — not as a hobbyist, but in production. I’ve led ML teams building personalized recommendation systems at enterprise scale, designed and shipped notification infrastructure used by millions of users, and built the experimentation frameworks (multi-armed bandits, A/B testing) that sit between a model and a decision. I hold a B.S. in Computer Science and an M.S. in Data Science with a capstone in Machine Learning.
The home lab isn’t where I learned this stuff. It’s where I keep learning it — with full control, no budget constraints from a ticket system, and no one to blame when something breaks.
What I Write About
Four areas, all overlapping:
Machine learning — From building and shipping recommendation models in production to running LLMs locally. I’ll write about what actually works — the infrastructure, the tradeoffs, the failure modes that don’t show up in research papers.
Software engineering — Design decisions, architecture patterns, the stuff that takes years to accumulate. Distributed systems, microservices, the decisions that matter long before you write a line of code.
System design — How to think about building systems that hold up under real load. I’ve designed APIs serving 15-20k RPM and pipelines processing large datasets with Kafka and Airflow. This is where I’ll get into the tradeoffs.
Home lab — Real hardware, real networking, real failures. This is my R&D environment. If you’re building your own infrastructure, you’ll find something useful here.
Who This Is For
Engineers who want to understand how things actually work — not tutorials that show the happy path. Whether you’re building ML systems at scale, designing distributed infrastructure, or just trying to own your own stack instead of renting it, this is for you.
The perspective here comes from having shipped this stuff in production. Not just theorized about it.
One Practical Note
This blog has no tracking, no ads, and no comment section. It loads fast and gets out of your way. I built it on Astro and deploy it to Cloudflare Pages — because simple infrastructure that works is kind of the whole point.
More posts coming. Let’s go.