DOCUMENT WORKFLOW STUDIO

The challenge

When every document needs to become one row, a skipped field or a value placed in the wrong column can compromise the entire spreadsheet. After several files, it also becomes difficult to return to the source and check what was copied.

Who it serves

I developed the prototype for people who regularly receive PDFs and spreadsheets and need to collect a specific set of information in a new Excel file without losing the ability to review the result.

What I built

I designed and developed the complete desktop application. I implemented file selection and parsing, field selection, reusable setup saving, the review table and .xlsx export. During review, every row keeps its source-file reference, and values the application cannot find remain marked.

How it works

I developed the application with Electron. PDF.js reads selectable text from PDFs, and ExcelJS reads and creates .xlsx files. The interface runs in a sandboxed process and uses a preload bridge for file selection, parsing and export. In this version, the parser recognizes lines in the “Label: Value” format and uses the first spreadsheet row as its header. There is no server or external AI service.

Key decisions

• The source files remain on the computer. The application only saves the reusable field setup; it does not create another stored copy of the documents or extracted records.
• When a value is not found, it appears as missing. The file, page, sheet or row reference stays beside the extracted value so the person reviewing it can check the source.
• For this first version, I used deterministic rules because it is possible to understand why each value was recognized. OCR, fuzzy matching and AI interpretation are not part of the current prototype.

Outcome