About

modelrisk.ch is an independent research project by Daniel Huber.

I lead AI infrastructure at a Swiss structured products issuer. This project grew out of a question that role keeps raising but cannot answer on employer time: what would it actually take to test, monitor, and explain a large language model well enough that a Swiss model risk manager could sign off on it? The work here is independent of that role. It uses:

All structured-product examples on this site are synthetic, based on public SSPA product structures. They are not real Swiss structured products in the regulatory sense.

This is a single-author project, done in personal time on personal infrastructure. Methods, calibration statistics, and limitations are documented so the work can be judged on its content.

Contact: danielhuber.dev@proton.me