A roughly fifty-person startup building Flux: a whole-head magnetoencephalography helmet made from 432 optically-pumped magnetometers, aimed at recording brain activity during natural human experience rather than inside the fixed geometry of a conventional MEG scanner.

The work spanned the whole instrument. Sensor assembly and calibration, custom CAD fixtures and manufacturing-process iteration alongside physicists and engineers, helmet integration and signal validation; the Python signal-processing and quality-assurance platform — preprocessing, artifact suppression, feature extraction, performance dashboards — whose metrics fed back into hardware design; and co-ownership of the acquisition environment, running participant studies that paired cognitive tasks with real-time MEG. Co-author on the system’s SPIE 2021 paper.

Two things came out of it besides the hardware. The first was a constraint, read correctly: clean MEG data requires a magnetically shielded room, which is a fundamental scaling problem for anything meant to be field-deployable. Kernel discontinued the Flux line not long after and moved to fNIRS. The second was a lesson about the field — scientists have a hard time building products, particularly in neuroscience. The work stays researchy, and companies raise substantial capital while struggling to ship. Closing that gap, in myself, is what the shipping cadence of the current work is for.