The language model is part of the tool, but it is not the whole method. Every review follows consistent criteria, requires evidence from the document and must not add experience that is not present in the CV.
ChatGPT can provide useful feedback. The problem is not a lack of model intelligence, but the absence of a repeatable process and controls over what counts as evidence.
We do not claim to have created a proprietary language model or to reproduce a real decision made by a particular recruiter. CV Unfold's value lies in its review process and the quality controls placed around an AI model.
We verify the file type and signature, extract text from a PDF or DOCX and reject documents that cannot be read safely.
We account for the selected specialisation, level and — optionally — a specific job description.
We assess seniority, ownership, impact, technical credibility, clarity and basic ATS hygiene.
A material finding must reference the CV. Missing information becomes a gap or a question, never an assumption.
We check the schema, evidence references, confidence levels and a deterministically calculated final score.
We do not ‘tune the model’ in the sense of training a proprietary AI. We continuously develop and version the system around it: prompts, rubrics, validation and the way findings are presented.
Controlled scenarios test evidence, factuality, prompt injection resistance, report language and the distinction between a gap and a missing skill.
We record the model, prompt and rubric version so that a result can be tied to a specific configuration and compared after a change.
We monitor error categories, analysis success, cost and voluntary feedback — without sending CV or report content to analytics.
A new model or material rubric change must pass the same tests and a review of representative outputs.
Rephrases existing information without adding facts.
Surfaces a fact already present in the document.
Marks evidence that is not visible for the target role.
Requests information required for a safe improvement.
Paweł has worked as a Frontend Developer, Senior Frontend Developer, Technical Lead and team leader. He specialises in React, frontend architecture and digital product quality. While leading technical interviews, he saw both strong experience presented too weakly and impressive claims without sufficient evidence. That distinction is the starting point of the CV Unfold methodology.
CV Unfold is Paweł's independent product and does not represent any current or former employer.The initial result is free. You only pay for the full report after seeing a specific diagnosis.
The source document is not stored in product storage.