Richard Landy

IMPEL began as a lab tool written by BrightSpot's founding scientists — one power supply, a consumer camera, single shots of small samples. Ambitions had outgrown the software.

Stabilize first, then re-architect: a multiprocessing pub/sub core of local services communicating over dedicated queues and shared memory, concurrent hardware control, and a CUDA-enabled ML pipeline — production-grade concurrency on ordinary line PCs, fully offline.

EL and PL imaging demand long exposures and tightly coordinated illumination — and LED power conditions drift, so without active control, luminous intensity wanders over the exposure and measurements stop being comparable.

Capture orchestration across an array of actively monitored power supplies, later matured with constant LED intensity control through long exposures — the difference between images you can look at and measurements you can trust.

Precise feature measurement across a field of view no single camera covers — and stitched imagery is only as trustworthy as the calibration behind it.

Multi-camera rig calibration underpinning image stitching and quantitative feature measurement.

Inline QC only counts if the inspection system speaks the factory's language — MES, PLCs, and the automation around the line.

Integrated IMPEL with the MES via PLC alongside REST API controls and a growing set of supported peripherals, making inspection a first-class step in the manufacturing flow.