Skip to content
RoleFramework
EN, Language: English
Log inCreate account

How it works

Four steps, from the resume you already have to an application you can trace back to the exact version you sent.

  1. Import the resume you already have

    Upload your resume as a PDF or a Word file, up to 10 MB and 10 pages, or paste a Google Docs link. From a PDF, RoleFramework reads the text, fonts and positions, rebuilds the page as editable code, renders it again and compares the two line by line.

    You see the original and the copy side by side, with one of three results:

    Exact
    The same pages, every line's text identical and within 1.5 points of where it was, with no substituted fonts.
    Faithful
    The same number of pages and no missing lines, every line within 2 points of where it was.
    Rebuilt
    The closest layout we could build. The differences are marked so you can decide.

    Today: PDFs printed from a browser import Exact. In our tests, PDFs from Word, LaTeX or Typst came back Exact or Faithful on 13 October 2026, and two-column resumes with a sidebar or a divider on 14 October; others may come back Rebuilt. A Word file is measured against LibreOffice's rendering of it; our sample Word resume came back Exact on 14 October 2026. A scanned resume is read with text recognition and labelled Rebuilt from a scan. A Google Docs link works when the document is shared with anyone who has the link.

    The import result for a synthetic resume: a readout saying Exact, 45 of 45 lines, largest shift 0.75 pt, all fonts matched, above the original PDF and the editable rebuild side by side.
    Original and copy side by side, measured.

  2. Confirm what is true

    Before anything is rewritten, you answer short questions about the work behind your bullets: the numbers, the scope, what you did yourself. Each answer becomes a fact you can read.

    A draft can use only facts you confirmed. If you skip a question, nothing is claimed from it.

    Question cards during tailoring. One asks about AWS and EC2 experience, says why it asks, and lists statements a yes would let the draft claim, each with a checkbox; on the card above, “Skip, do not claim” is ticked.
    Tick what is true, or skip.

  3. Tailor to the posting

    Paste a posting, or a link to it. First you get a fit rating with reasons, your level against the role and the languages it asks for, so you can decide whether to apply before you spend time writing.

    Then you get a draft. Each changed line shows the facts that support it, and the page limit is drawn on the sheet so you can see what fits. You review every change before it becomes a version.

    Drafts are written with the help of a language model. Your name, contact details, links and photo are removed before anything is sent to it.

    Today: salary and interview research come from a web search, with each source listed and dated, and are labelled “Estimate” when nothing specific to the role is found. Postings in Swedish and English are read, and “Make a Swedish variant” or “Make an English variant” in the editor saves the version in the other language as a new version. LinkedIn job links are fetched when LinkedIn allows it; when it blocks the fetch, paste the text and the link stays attached.

    The assessment of a fictional Platform Engineer posting: fit 8 out of 10 with the advice Apply, the reasons, and a table that sets each requirement against the evidence in the resume.
    The fit rating comes first.
    A draft for a fictional posting: four changes numbered on the sheet and listed beside it, three reworded lines each marked “Supported by 1 fact you confirmed” and one moved line, with the dashed red page-limit line and 3.1 lines of room left.
    Each change shows the fact behind it, and the room left.

  4. Track the outcome

    When you apply, mark it. The posting, the exact version you sent and each interview round stay together, and every application is on one page, grouped by stage.

    Add reminders to an application: follow up, the application deadline, a thank-you note, preparing for the interview. Each application shows its next reminder.

    Already keep a tracker in a spreadsheet? Import it as a CSV file, and export yours the same way.

    The applications page with four applications at fictional companies, counted by stage at the top and grouped below: two under Applied, one under Recruiter screen and one under Interviewing. Each row shows the role, the date you applied and the resume version you sent; one has a follow-up due today.
    Applications by stage.

Before you export

Seven checks run before you export. If one fails, you see what is wrong and how to fix it, and you decide whether to export anyway.

  1. Page count

    The version has no more pages than the resume you imported.

  2. Neutral file name

    The file and its PDF title are named after you, with “Resume” (or “CV” for a Swedish resume), never after the company.

  3. No company name

    The company you are applying to is not named anywhere in the text.

  4. No non-breaking spaces

    Words stay separate when the text is copied or read by software.

  5. Readable text

    Every word software extracts from the PDF is visible on the page. No hidden text.

  6. Claims confirmed

    Every changed line has a receipt: a fact you confirmed.

  7. Every line of your file

    The import behind the version has every line of the file you uploaded. If some are missing, the check counts them and the compare view lists them.

ATS-friendly exports

PDF exports are text-based: every line is real, selectable text in embedded fonts, and icons are decorative. Reading order is kept column by column. The open-source PDF parsers poppler, PyMuPDF and pdfminer read them back as completely as the original, or better.

A Word export keeps your file's heading styles and reads back in order. A scanned resume has no text a parser can read; its export carries the recognised text as real text.

Tested on 20 October 2026 with 11 resumes: ten made in a browser, Word, LaTeX, Typst or Pages, and one scan. Each was exported as imported, and again in the layout a tailored version uses, with three edits on all but one. PyMuPDF and pdfminer found every word of every export, and poppler as many as in the original or more. We tested these parsers, not any applicant tracking system itself, so there is no ATS score.

Start with the resume you already have.

Import your resume

Free during the beta, then a paid subscription. No payment details are asked, and you hear the price before the beta ends.

How it works | RoleFramework