Public Goods: Selected GitHub Repositories

Here are some of the projects I’ve worked on that I’d like to showcase.

Historical Population Census 1931: Jati Counts across Granular Regions (2026)

Transcribes caste (jati) population counts from 26 scanned provincial and princely-state volumes of the 1931 Census (Imperial Table XVII) into a labelled Stata panel: 69,305 rows across 49 provinces/states, 1,849 caste names and 1,200 districts, towns and tehsils, with a generated codebook and validation report.

• Computer-vision page segmentation (OpenCV ink-projection profiles, Tesseract deskewing) cuts each table into tiles along its own column boundaries before any model call.
• Two independent vision-model readings per tile via the Anthropic Batch API, validated with column-alignment, sex-sum and geographic-hierarchy checks.

Tools: Python (OpenCV, pdfplumber, NumPy, pandas), Anthropic Batch API, Poppler, Stata

Repositories

The repository data below is fetched live from GitHub!