decided to write a script to see what the actual text-extraction layer sees when it reads my PDF.
the results were pretty bad:
Column Bleed: The parser was reading my "Skills" sidebar line-by-line along with my "Experience" section. It turned "React" into a middle name for my previous boss.
Icon Failure: The little email/phone icons I used instead of text labels resulted in null contact info fields.
i built a simple next.js wrapper to visualize the raw JSON output so i could "debug" my resume format. fixed the layout, parsed it again until the JSON was clean.
posted the findings and the tool on reddit yesterday. it hit 52k views and #1 trending in the sub. woke up today to a permanent account ban. apparently showing candidates the raw parsing data is "spam" or "manipulation."
anyway, if you want to check your own resume's parsing score, i kept the site up:
https://resumefy.pro
(it's paid now to cover the server bill, but the logic is simple: keep it single-column and avoid tables).
0 comments