I build cloud backends that have to stay up — and small sharp tools for the things that annoy me on the way there.

Quantum computing from first principles to AWS Braket — where the cat is in a superposition of understands quantum computing and does not, until you measure it with the exercises.

An offline-first e-learning platform and full curriculum: eight terms that start at classical mechanics and end at running circuits on AWS Braket. Built as a zoneless, prerendered Angular PWA with KaTeX for the maths, Python for the worked examples, and Leitner spaced-repetition flashcards so the material actually sticks.

Do the free-text comments astronomers write on TESS Objects of Interest carry disposition-relevant information that the numeric catalogue columns do not?

A language model turns ExoFOP comments into numeric semantic features; gradient boosting is the predictor; grouped cross-validation by host star is the arbiter. The model's own probabilities are never thresholded into a decision — astrophysics is maximally out of distribution for it. A null result is a valid outcome, and the repo is built to report one honestly.