Tools, systems, and experiments — mostly born from the question "what if I could automate this?"
A desktop application that automates the creation of HathiTrust-compliant digital preservation packages from TIFF images. Handles OCR processing via Tesseract, generates compliant YAML metadata, assembles packages with checksums, and validates against HathiTrust requirements. Built for production use at Purdue University Libraries.
Parses Golden Thread NXT XML reports to extract FADGI (Federal Agencies Digital Guidelines Initiative) compliance data and tracks it over time in formatted Excel spreadsheets. Supports single-file and batch processing with automated trend analysis and summary dashboards.
A lightweight GUI tool that calculates optimal AutoCrop visual padding using the geometric mean of object dimensions. Supports loose and bound materials with live calculation as you type. Zero dependencies beyond Python 3 — just double-click and go.
A streamlined toolkit for transforming academic and historical PDFs into accessibility-ready documents. Generates OCR text layers for screen readers, strips orphan tags for Acrobat compatibility, enforces PDF/UA-1 metadata compliance, and supports batch processing with folder preservation. Used by Purdue Libraries' scholarly publishing division. Includes a scraper for downloading academic papers from digital repositories.
A chess-themed multi-agent system with specialized AI roles — orchestration, deep research, systems automation, utility tasks, and independent evaluation. Each piece runs a different model chosen for the right balance of capability and cost.