The gaps between what existing technology can do and what the world actually needs.
Novel Systems Engineering is Samuel C. Ryan's independent research workshop, focused on new problems in a changing world.
Two technology threads have come out of it: a polarimetric LiDAR classification architecture (DDR) and a multi-agent framework for mitigating sycophancy in language models (CANDOR). Both are patent-pending or published and available for review.
DDR classifies targets by material composition. Shape and motion are not factors.
When a pulsed SWIR laser illuminates a surface, the return signal depolarizes at a rate that depends on material structure. The asymmetry between horizontal and vertical polarization states is a physically grounded material signature, stable across target geometries and viewing angles.
Read more →
RLHF-trained language models tend to agree with whoever is talking to them. CANDOR mitigates this through structured agent debate.
The core mechanism, Principled Agent Debate (PAD), pairs two models as philosophical opponents (one for the user's position, one against) while a third evaluates the exchange. The white paper reports up to 34.5% accuracy improvement over control baselines.
Read more →
Samuel C. Ryan is a systems engineer and inventor based in Wilsonville, Oregon. He holds an M.S. in Semiconductor Process Engineering from the University of Oregon and has designed instrumentation systems across seven industries over fifteen years, including work at FLIR Systems on DOD-qualified thermal imaging hardware and at ESS Tech on long-term energy storage.
Novel Systems Engineering LLC is his independent research vehicle, focused on problems that don't yet have a commercial solution.