How Keyverta works and who is behind it

Who runs Keyverta

Keyverta (keyverta.com) is a service operated by OnruTech, which is responsible for developing the product, running it and publishing the guides. OnruTech is also the data controller described in the privacy policy.

Contact: onrutech@gmail.com

How an analysis is prepared

When you compare your resume with a job posting, the process follows these steps:

1. Text extraction. We read the text of the PDF or DOCX you upload; the original file is not kept, only its text. You can review the extracted text. We do not apply optical character recognition (OCR): a resume scanned as an image may contain no usable text.
2. Data minimisation. Before sending the text to the AI model, pattern matching masks email addresses, phone numbers and links. This is minimisation, not anonymisation: names, employers and other details remain.
3. Comparison with the job. A large language model (LLM), run through the provider Groq, compares your resume with the job description and returns a structured report: score, job title fit, missing keywords, skills, suggested rewrites and format warnings. Its instructions tell it not to invent experience and to phrase achievements as templates you fill in.
4. Cross-check against your text. If the model flags a keyword as missing that does appear in your resume (ignoring case and accents), we drop it. A second request to the model removes keywords you already cover with different wording.
5. Format checks. Fixed rules run on the extracted text to detect signs of tables, multiple columns, decorative symbols, a missing email or phone number and very short text. They are added to the model's own warnings.

What the analysis is not

• It is not the employer's ATS. Each employer uses its own software and configuration; Keyverta has no access to them and does not replicate them.
• The 0–100 score is guidance produced by the model from the two texts. It is not a hiring probability or an official grade, and it can vary between analyses.
• It can be wrong: the model may miss a requirement, misread a sentence or suggest wording that does not represent you. Review every suggestion before using it.
• It does not guarantee interviews or job offers. Those decisions rest with the employer and depend on factors the document does not capture.

How we prepare guides and examples

Guides are published under OnruTech's editorial responsibility. Resume models, job postings and demonstration comparisons are fictional cases labelled as such. They are not testimonials, hiring outcomes or user results.

We prioritise examples you can adapt, instructions linked to real features and references when a claim depends on a source. Advice is not universal: adapt your document to the country, sector and application instructions. A review date indicates a content review, not approval by an external recruiter.

Availability and data

Accounts and tools are currently free. Analysis requires registration and is limited to five analyses per user per day, alongside rate and capacity limits. Generating a tailored resume or cover letter depends on the availability of the AI service. Fixed response times are not guaranteed.

The builder draft is stored only in your browser. When you analyse a resume, its text and the job posting are stored in your account and sent to Groq to generate the analysis. You can delete individual analyses or delete your account with all its content. The privacy policy details providers, retention periods and your rights.