About
I build software that has to be fast, and I enjoy the part where you find out why it isn't.
Most of my work sits close to the machine: systems programming, performance analysis, and compilers. I like problems where the answer is measurable — where you can point at a number before and a number after and argue about what happened in between.
What I'm working on
Lately: SIMD-optimised dataframe internals in C++, prediction-market machinery in Rust, and a browser-resident JIT that exists purely because it seemed like a funny thing to build.
How I work
I'm a believer in cheap experiments with pre-registered predictions. Writing down what you expect to happen — with an actual number — before you run the benchmark is the difference between learning something and confirming what you already thought.