Collect the original project context
Open your career notes, repository records, project documentation, or other source material and gather the facts that existed before any AI rewrite. Record the project name, dates, ownership, purpose, collaborators, payment status, intended audience, operating environment, and evidence for each outcome. Separate what you personally did from what a library, school, community, or other contributor did. If a fact is missing, mark it as unknown instead of supplying a plausible answer. This first pass establishes provenance and gives the generator a factual basis for distinguishing personal work from employment.
Add an explicit relationship label
Give every project a clear relationship label before asking for resume language. Suitable labels include personal project, open-source contribution, volunteer work, academic work, contract work, or employed work. Add a short explanation where the label could be misunderstood, such as “personal project published publicly” or “volunteer contribution for a community group.” Do not rely on a project title, repository visibility, or technical complexity to communicate the relationship. A direct label makes it harder for a model to convert an interesting project into an invented job, client engagement, or employer assignment.
Audit claims about use and operations
Review every statement that could imply real customers, business ownership, production operation, or organisational responsibility. Ask whether the project had confirmed external users, what data it handled, where it ran, and whether anyone depended on it. Distinguish a local test, a public demo, a hosted preview, and a production service. If your evidence only shows that the application worked in a test environment, retain that narrower description. Avoid turning a deployment method into proof of operational use. This is the point where unsupported employment narratives often enter an otherwise accurate project description.
Separate activities from employment signals
Rewrite source notes so they describe actions without borrowing the language of a job unless a job actually existed. “Built,” “documented,” “tested,” and “contributed” can describe personal or community work. Terms such as employer, client, team assignment, paid engagement, company product, and production owner require supporting context. Also check titles: calling yourself a founder, consultant, or engineer may be misleading if the title was never used or if it suggests a relationship that did not exist. Preserve the useful technical detail while removing status claims that the source notes cannot support.
Generate with uncertainty preserved
Give the cleaned notes to the resume system and instruct it to use only confirmed context, report gaps, and ask questions rather than inventing answers. The Radaryno Resume Kit uses buyer career notes, selects relevant experience, reports gaps, and asks rather than inventing; its generated files stay on the buyer computer. It can produce an HTML and PDF resume in the offer language, but output format does not validate a claim. Review whether the generated entry keeps the relationship label, dates, audience, and operating boundary intact. If the system asks about an unsupported detail, leave it unresolved until you verify it.
Compare the draft with the source record
Perform a final provenance comparison line by line. Confirm that every employer, client, title, payment statement, user claim, and production reference appears in the source notes or has been separately verified. Check that personal and open-source work remains in an appropriate section and is not merged with employment history merely because the wording sounds professional. If the draft is too vague, add documented scope or technical detail; if it is too strong, narrow the wording. Keep a corrected version of the source note so later generations do not reintroduce the same unsupported context.
| Source context | Payment or employer | Safe description | Claim to avoid | Required evidence |
|---|---|---|---|---|
| Personal project | No employer or client | Built and maintained a personal tool to explore a stated problem | Presented it as company work or client delivery | The notes confirm the build, maintenance and personal setting. |
| Open-source contribution | No employer unless separately documented | Contributed code, documentation, or issue work to a public project | Called the project a job or paid engagement | The notes identify the actual contribution and relationship. |
| Volunteer work | An organisation may exist; payment is not assumed | Created or improved something for a named organisation as a volunteer | Implied an employment relationship | The notes confirm volunteer status and the described activity. |
| Academic work | School or research setting | Completed a course, thesis, or research project with defined scope | Claimed commercial ownership or customer use | The notes confirm the academic setting and completed work. |
| Contract or employment | Documented client or employer | Reported work under the confirmed role and dates | Added an unverified title, client, or operational outcome | The notes confirm the relationship, title, dates and actual work. |
A worked provenance review
Consider a fictional developer who built a budgeting application at home, published the source code on GitHub, and received no payment. The initial note says, “Created a budgeting platform used by customers.” That wording combines a personal project with an unsupported customer claim, so it is not safe input for a resume generator.
The developer checks the evidence and records a narrower context: personal project, unpaid, self-directed, published repository, demonstration data, and no confirmed external users. The work may still show technical decisions, testing, documentation, or deployment practice, but those claims must remain tied to what actually happened.
The corrected source note becomes: “Built and maintained a personal budgeting application with a JavaScript interface and PostgreSQL storage; published the code on GitHub; tested with sample data; no employer, client, or confirmed production users.” Each clause identifies an observable activity without implying employment.
A generated entry can now describe the build, technologies, and testing boundary. If the system asks whether the application supported customers or operated in production, the answer should remain unknown or no rather than being inferred from a public URL. The project is evidence of personal work, not employment.