Projects

Reproducibility repo for the Dorabella paper. Ships the matched-budget simulated-annealing scripts that produce the §4 calibration result (z = −1.23, below the shuffle mean), the substitution-consistency check that falsifies Packwood (2020), pre-built English/Italian quadgram models, and the cipher transcription (Hauer 2021).

Python library and CLI for calibrating dictionary hit rates. Given a list of short strings and a reference dictionary, separates real matches from chance collisions. Applications include decipherment, OCR validation, spell-check evaluation, password audits, and fuzzer QA.

Research paper quantifying the noise introduced when computational decipherment checks decoded tokens against a large dictionary. Develops a four-category classifier that separates real signal from dictionary-induced collisions, validated across 252 experimental cells against five competing correction methods.

Data pipeline cross-referencing WARN layoffs, H-1B petitions, SEC filings, and quarterly financials for S&P 500 companies. 308 layoff events, 607K visa records, 29K WARN notices, one SQLite database.

Syllabary analysis pipeline with signal isolation: 77-phase computational analysis identifying Italian syllabic tachygraphy as the Voynich Manuscript's encoding mechanism.

Letter-level pipeline matching consonant skeletons against a Latin medical dictionary, the first of two independent approaches to the Voynich Manuscript.