Coverage · Editorial map
Six standing subject areas anchor the desk’s reading, review and critique of AI in bond-portfolio risk assessment.
Machine-learning approaches that re-price credit curves and score issuer-level default risk. Reviews ask how such models behave with illiquid issuers, thin default histories and abrupt spread repricing.
Scenario engines that generate yield-curve and macro paths for stressing duration and convexity — including the foreign-currency books hedged back to New Taiwan dollars that local institutions manage.
Models that flag credit deterioration before agencies act. The desk examines what counts as a signal, versus what merely counts the calendar of rating announcements.
Analytics that estimate how much a corporate-bond book can move without paying for the privilege — often the least forgiving test an AI method faces.
The quieter half of the subject: how methods are tested before portfolios rely on them, and whether their explanations inform decisions or merely accompany them.
Behind every confident model sits reference data, curve construction choices and vendor claims — the desk reads these the way mechanics read a service manual.
A method you want reviewed that fits none of these areas? Ask the desk directly — send an inquiry and it will be considered on its merits.
.jpg){kind=link}