Pipeline: the processing steps

Notebooks for data processing and analysis (notebooks) are grouped into task types:

Stage	Description
01_setup	Define folder structure, set preferences, lock administrative referencing
02_ingest	Download, unzip, and stage data in cache
03_harmonize	Align entity datasets from multiple sources, create spine of entities
04_enrich	Produce entity-keyed evidence from imagery, models, geoprocessing, or record linkage without selecting canonical values.
05_curate	Create canonical datasets by reconciling evidence, imputing and inferring attributes, and filtering records.
06_model	Fit models, cross-validate, derive standard errors, optimize.
07_infer	Create artifacts from models: predictions, aggregated coefficients
08_report	Creation of publication-ready figures, tables, and text
09_show	Interactive display of results, demonstrating quality, issues, or functionality

These notebooks can be converted to scripts and be orchestrated across large computing clusters (to process data for entire world regions).