Thinkhome

Suppliers send product catalogs as PDFs. Getting those into a browsable, searchable store is normally weeks of copy-paste. This pipeline does the extraction and leaves humans to do the part they are actually good at: checking it.

Extraction in parallel

An uploaded catalog is processed on two axes at once, page by page and product by product, so a large document does not become a long serial job. The output is structured data rather than a wall of text.

Review before publish

Administrators get a PDF viewer with separate page and product views, so they can check an extracted product against the page it came from without leaving the screen. Extracted fields are editable, and nothing reaches the storefront until it passes review.

Approved content publishes through a prebuilt static frontend, which keeps the customer-facing site fast while the admin side stays dynamic.

Working in an existing codebase

The Next.js and Strapi setup was already there when I arrived, so alongside the pipeline I worked across both, updating and cleaning them up to meet new requirements instead of layering more patches onto the old structure.