Systems at scale
A sharded, synchronous text-matching system at Meta, handling millions of QPS at millisecond latency, globally. The engineering of moving real traffic without falling over, and the performance work that makes it economical.
Software & systems engineering
Building software simply, so it can keep changing.
Software and systems engineer with 15+ years across startups, open source, and Big Tech, lately hands-on with AI agents and local models. I’m also a father of three, which taught me to listen before I solve. More about me on LinkedIn.
I advise early-stage teams on architecture, scaling, and AI strategy. Get in touch if that’d be useful.
Open-source frameworks
Core framework work on the Quarkus team at Red Hat, building the substrate other developers build on, where fast and efficient is the whole premise.
Agents from scratch
I built miniswe, a coding agent written from the ground up, to understand agents at the engineering level, not as a framework black box.
See minisweWriting
All posts- Making miniswe 2.5× faster, the data-driven way What it took to make my AI coding agent competitive with OpenCode: measure, get refuted, revert, repeat
- miniswe: A Lightweight Coding Agent for Local LLMs A lightweight, Rust coding agent built for small local LLMs
- Running Thorntail apps in the IDE with Thorntail Runner A cleaner way to run Thorntail applications from your IDE — using the JBoss Modules classpath instead of a flat one, so in-IDE runs match java -jar.