normix
JAX package for normal variance mixture distributions (generalized hyperbolic and relatives), with fitting and risk measures.
JAX · probability
Hi, I'm Xiang. I used to be a math PhD student trying to figure out why financial returns refuse to be Gaussian. These days I build risk models at a hedge fund and write code for fun.
This is where I keep my open-source code, math notes, and the occasional essay.
JAX package for normal variance mixture distributions (generalized hyperbolic and relatives), with fitting and risk measures.
Notes on fat tails, information geometry, and distribution theory.
A critique of Harari’s account of shared stories and large-scale cooperation.
Five common AI bubble arguments, from valuations to workplace productivity.
A math-class analogy for how three training approaches score an LLM’s answers.
Find my code on GitHub and my posts on X.