About
James Zoryk
Bayesian statistics, machine learning, probabilistic programming, quantitative finance and risk modelling.
- Bayesian inference and MCMC diagnostics — HMC, NUTS, reparameterisation, and the failure modes that don't make it into the textbook.
- Probabilistic programming — Stan, PyMC, NumPyro — and where the abstractions leak.
- Quantitative finance and risk — backtesting, multiple comparisons, and the ways a Sharpe ratio lies to you.
- Machine learning where the point is calibration and uncertainty, not leaderboard position.
Stan, PyMC, NumPyro, Python, R.
This is not a tutorials blog. Assume familiarity with probability and at least one probabilistic programming language. I write something down when I've had to solve it myself, not because the topic is trending — most posts here started as a script I needed for something else: a backtest, a hierarchical model that wouldn't converge, a derivation I had to redo because I didn't trust the first pass.
The stack this site runs on is boring on purpose: Markdown in, SQLite out, no build step between publishing a post and it being live. If something looks wrong in a post — a typo, a bad prior, a derivation that doesn't hold — the fastest way to reach me is the address above.