Your experience may fit a role even when your CV makes the reader work to see it. You want clearer, more relevant wording without sending the document to a cloud AI service.
OGAD (Off Grid AI Desktop) can compare your CV with a saved job description using a local model on your computer. Ask it to identify relevant evidence, suggest edits, and flag gaps. You choose the changes and verify every claim before applying.
| Download OGAD | Desktop releases |
What would you like to do with Off Grid AI?
Have a feature or use case you would like us to support? Tell us what you want to do and which device you use.
Write to support@offgridmobileai.co, join our Slack community, or talk to us on Reddit.
The useful result is a CV that explains your fit more clearly. It should still be your record of work, with the same employers, dates, qualifications, and outcomes you can support.
What does tailoring a CV involve?
Tailoring means selecting and explaining relevant experience for a particular role. It can change the order of examples, the emphasis in a summary, or the wording of a bullet. It should not invent a qualification or turn limited exposure into expertise.
Suppose you are a project coordinator applying for a delivery role. The job description asks for stakeholder communication, risk tracking, and cross-team planning. Your CV describes meeting scheduling and status reports but gives little context about why they mattered.
A useful review asks what evidence you have for each requirement:
| Requirement | Evidence to look for in your own history |
|---|---|
| Stakeholder communication | Who you kept informed and what decisions that supported |
| Risk tracking | What you monitored and how you raised a problem |
| Cross-team planning | Which teams you coordinated and what you were responsible for |
If the evidence is missing from the CV, the model should ask you for it. A plausible example is not a substitute for experience you actually had.
What do you need before you start?
Install OGAD and download a local text model that fits your computer. Save the job description as a readable local file. Complete the app and model downloads while connected, then use local inference for the comparison.
The procedure uses the free core desktop app on supported Mac and Windows computers. Release 0.0.54-beta.108 also provides Linux packages as a beta.
Use TXT, Markdown, DOCX, or a PDF with extractable text. If the PDF is a scan, create and check a text version first. Keep an untouched copy of your CV and edit a separate application version.
You can remove contact details from the working copy if they are not needed for the task. The model needs the experience and requirements to compare them; it does not need your phone number to improve a bullet.
How do you compare the documents locally?
Attach the CV and job description to a new chat with a downloaded local text model selected. Read the extracted text before asking for edits, especially if the CV uses columns or unusual formatting.
- Open Models > Text and select your downloaded model.
- Start a new chat and select + > Attach files.
- Add the CV and saved job description.
- Wait for processing and open the attachment previews.
- Check that job titles, dates, and bullet text were extracted in a readable order.
The desktop extraction code supplies document text to the model. It does not guarantee that your original layout or every visual relationship survives extraction.
Ask for a comparison before a rewrite:
Compare my CV with this job description. Create a table of stated role requirements, relevant evidence already in my CV, and gaps that need my input. Quote or name the CV passage supporting each match. Do not invent experience, skills, metrics, or qualifications.
Check the table. A model may match similar words even when the actual responsibility differs.
How do you improve a bullet without inventing results?
Give the model the facts you can support: your action, responsibility, context, and result where known. Ask for clearer language using only those facts. If a number is unavailable, leave it out.
For example, you might provide:
I collected weekly status updates from design and development, identified missing decisions, and prepared the agenda for the delivery meeting. I did not own the project budget. I do not have a measured time-saving figure.
Then ask:
Suggest two concise CV bullets from these facts for the delivery role. Preserve my actual level of responsibility. Do not add budget ownership, team size, or quantified results.
Choose wording you could explain in an interview. If the sentence sounds more senior than the work you did, revise it. A clear account of a real contribution is more useful than a claim you cannot defend.
How do you choose what to emphasise?
Prioritise the strongest supported examples for the role. Ask the model to explain the reason for each suggested change, then decide whether the role description and your experience justify it.
A useful request is:
Suggest which existing bullets should appear earlier for this role. For each suggestion, name the requirement it supports. Keep important career context and do not remove information merely because the job description uses different wording.
You can also revise the opening summary after the evidence review. Ask for a brief summary based only on the verified matches. Avoid broad labels such as “expert” unless your record supports them.
Do not repeat every phrase from the job description. Use the employer’s terms when they accurately describe your work, and keep ordinary language where it is clearer.
How should you handle a requirement you do not meet?
Mark it as a gap. Decide whether you have related experience worth explaining, whether you should learn more, or whether the role is a poor fit. The local model can help frame a question, but it cannot supply missing experience.
For a tool you have only observed, say so accurately. For a responsibility you supported rather than owned, keep that distinction. Ask:
Which proposed edits overstate my experience compared with the original CV and facts I supplied? Show the original and suggested wording side by side.
Review the answer yourself. Asking the model to audit its wording is useful, but it is not an independent verification of your employment history.
How do you check the final application version?
Compare the edited CV with the original and your own records. Check dates, role names, qualifications, metrics, and claims about ownership. Then read the document without the job description beside it: it should still make sense as your career history.
| Check | What to confirm |
|---|---|
| Accuracy | Every factual claim is supportable |
| Relevance | Important evidence is easy to find |
| Clarity | Bullets state what you did |
| Consistency | Dates and titles agree across sections |
| Layout | The final file is readable in the format you will submit |
Copy approved text into your normal document editor. Inspect the exported file before submitting it. This workflow suggests wording; it does not guarantee hiring outcomes, screening scores, or employer responses.
Try one role and one section
Download OGAD, attach a working CV and one job description, and review the evidence table. Rewrite one section using facts you can explain in an interview.
Keep the selected model local for this comparison. Initial downloads need a connection, and submitting the final application is a separate online step. The point is to keep AI drafting on your computer while improving how clearly your experience is presented.