A platform that turns photographs into printed ID cards

The problem

An ID card programme had complete records for every person already, so the only genuinely new input was the photograph. The existing process had operators re-entering data that was already on file, which made each record slow to produce and introduced mistakes into records that had been correct to begin with.

What we built

• Bulk import of existing records, with duplicate detection, warning prompts and recoverable deletes backed by database level audit triggers
• A capture flow reduced to the one thing that was actually missing, after timing analysis showed data entry, not the operators, was the bottleneck
• Operator accounts with number verification, per location assignment and activity monitoring, so supervisors can see what was captured and where it stalled
• An extraction and matching pipeline: photographs are matched back to the right record automatically, with anything uncertain sent to a human review queue
• Automatic card generation at press ready size with correct bleed and safe margins, packaged one location per PDF with spreadsheets and photo archives
• Self hosted image processing for background removal, replacing a paid per image service at zero running cost

Where it stands

A working pipeline from raw photograph to print ready PDF, with the card format approved by the press after it was matched to real measurements rather than to a screen design.

Worth knowing

The most valuable work was measurement. Timing every step and counting the real failure rate changed the capture method completely, and saved rebuilding the wrong thing.