Skip to content
CVForge

Product · Updated 2026-09-28

How CVForge works

CVForge is not a single instruction to 'improve this CV'. It is a pipeline where each stage produces structured data that the next stage consumes, and where the facts extracted from your document stay separate from the wording generated around them. That separation is what makes the factual check possible.

The pipeline

  1. Upload. Your CV or resume arrives as a PDF or DOCX, up to 10 MB.
  2. Extraction. The file is parsed into text. PDFs are limited to a readable page count and every parse is time-boxed.
  3. Section detection. The text is split into contact details, summary, experience, education, skills, projects and certifications.
  4. Fact extraction. Concrete facts — employers, titles, dates, qualifications, technologies, numbers — are pulled out into a structured record and kept separately from anything generated.
  5. Quality analysis. Each section is assessed for vague wording, missing outcome evidence, unclear scope, unnecessary content and structural problems.
  6. ATS analysis. The document is checked for the layout and formatting choices that break automated parsing.
  7. Job description analysis. If you supplied a posting, its requirements are extracted, split into must-have and preferred, and compared against your document.
  8. Improvement generation. A structured rewrite is produced that works only from the extracted facts.
  9. Fact verification. The generated content is checked against the extracted facts. Anything unsupported is rejected.
  10. Reconstruction. The result is rendered as a single-column, text-based PDF within the page limit you selected.

Why it is staged rather than one-shot

A single prompt that reads a CV and returns a better CV has no way to know what was true in the original. It has only the text and the instruction. Everything it produces that looks like a fact but was not derived from the document is indistinguishable from a real one, and nothing catches it.

Extracting facts first and generating around them changes the problem. The generator is no longer asked to both decide what matters and write it — it is asked to express, and it works from a fixed record. When a section of the original has no supporting evidence, that absence is visible in the record rather than hidden in the output.

How page limits are enforced

The selected plan determines the page budget. When the rendered document would exceed it, the renderer compresses typography and spacing within defined limits and then re-counts the pages. If the content genuinely cannot fit, CVForge reports that rather than silently trimming your experience to make the number work.

Frequently asked questions

How long does it take?
Generation runs in the background and the status page updates as each stage completes.
What happens to my file?
It is stored in a private bucket and deleted 12 hours after the order is created, whether or not it was paid for. It cannot be recovered after that.
Do I need an account?
No. Checkout is guest-only and no account is created.

Improve the CV you already have

Upload a PDF or DOCX and CVForge returns a sharper, ATS-safe, page-limited PDF. Your facts are preserved; gaps are reported rather than invented.

Start with your CVHow it works

CVForge is an automated CV enhancement tool built by Januth Nimnal as part of OrinAI. It improves an existing CV and does not write one from scratch. About CVForge.