What

Kernel Flux: a whole-head magnetoencephalography system built from 432 optically-pumped magnetometers. Conventional MEG uses cryogenically cooled SQUID sensors fixed in a one-size helmet; optically-pumped sensors are small and run warm, so they can sit close to the scalp and tolerate natural head motion. The goal was SQUID-grade signal recorded during natural human experience.

The work ran from early prototype to lab-ready instrument: sensor assembly and calibration, custom fixtures designed in CAD, iteration on the manufacturing process itself alongside physicists and engineers, and then integration of the sensors into the helmet and validation that the array actually recorded brain activity.

Why it mattered

A new sensor modality only becomes an instrument once it can be built repeatably and you know, quantitatively, how good each channel is. Nothing downstream — no pipeline, no study, no paper — is worth anything until that loop between manufacturing, measurement, and design is closed. It is also the most literal version of the belief that runs through this timeline: if the data is the bottleneck, go build the thing that collects it.

My role

I hand-assembled the first sensor units and tested them on my own brain.

Concretely: assembly and calibration of the array from prototype to lab-ready; custom CAD fixtures for the assembly process; process iteration with the physicists and engineers who designed the sensors; helmet integration; and signal validation — most often on myself, checking for a clean alpha rhythm as the fastest honest test that a channel was working.

Outcome

Co-author on the SPIE 2021 paper describing the completed system, Kernel Flux: a whole-head 432-magnetometer optically-pumped magnetoencephalography (OP-MEG) system for brain activity imaging during natural human experiences. Kernel has since discontinued the Flux line and moved to fNIRS — this stands as a hardware credential, not a live product.