01
The challenge
Spare-parts catalogs were spread across huge PDFs with inconsistent part-number formats, and some were OCR-scanned with noisy text, making manual lookup slow and error-prone.
02
The solution
We built an indexing engine that extracts parts and descriptions with source-specific parsing, normalizes part-number formats, corrects common OCR issues, and powers a desktop search UI that opens the original diagram or table directly from the PDF.
03
The result
The result is one searchable system indexing 15,274 parts across 3 catalogs, cutting part lookup time from minutes of manual browsing to a few seconds.