Apr 12, 2026Engineering
How we shaved 40% off our solver runtime
A deep dive into vectorisation, caching and where the real bottleneck lived (it was not where we thought).
Engineering write-ups, research notes and product updates. Written by the people building the platform — no ghostwriting, no marketing pass.

A deep dive into vectorisation, caching and where the real bottleneck lived (it was not where we thought).
What worked, what we threw away, and the one decision we would re-make today.
The counter-intuitive cases where exhaustive search outperforms our genetic optimiser for coil geometry.
Modelling parasitic effects at MHz frequencies — why the textbook approximation fails above 6 MHz.