Case Studies / Document Workflow Lab
A document workspace for people who need to merge, redact, or make PDFs searchable while understanding where processing happens and being able to inspect the exported result.
01 / Problem
Routine PDF work should make document processing location and downloadable results clear.
02 / Contribution
Supported operations run in the browser, so the document never leaves the device; operations that need the server use an explicit, visible server path. Every example below was checked by reopening the exported file, not just the screen.
Merge, redaction, and OCR were run on synthetic documents, and each exported file was reopened and checked. The captures and files below show the results.
Verified synthetic workflow evidence
Evidence package accepted on 15 September 2026. It contains synthetic documents only; captures are Chromium screenshots, not production or physical-device proof.
Two synthetic PDFs are queued, merged, then inspected again as a downloaded PDF. The 390px capture shows the same populated workflow at a narrow viewport.
Processing boundary
Browser-local document processing. The captures show real candidate downloads generated from synthetic inputs.
A synthetic marker is selected, then the resulting PDF is reopened after the server raster-rebuild redaction flow.
Processing boundary
The selection image is not proof of removal. The output package verified that original text and PDF objects were removed; raster rebuilding loses the original text, forms, and vector content.
An image-only synthetic English scan is used for browser OCR text extraction and for a separately inspected server-generated searchable PDF.
Processing boundary
Browser OCR used an explicitly preloaded local English model. The searchable PDF is server output for this simple English synthetic document; this does not claim multilingual, default offline, client-document, or production behavior.
I can build a practical document tool with a clear split between browser processing, model downloads, and backend processing, backed by a large automated test suite.