Engineering
How the on-device resume engine works
The generation engine in PocketResume 10.7.0 runs locally. That changes the failure modes, the cost model, and the privacy story. Here is how it works.
· 4 min read · By Isaac Daniel
Profile in, structured resume out
You add a master profile once: contact details, roles, dates, bullets, skills, education, and projects. PocketResume stores that as structured data. When you generate, the engine selects the relevant sections, orders them for the chosen template, and produces a resume document.
The job description is optional
If you paste a job description, the engine extracts terms and uses them to bias emphasis: which bullets rise, which skills are surfaced, and which language mirrors the posting. It does not create new employers, titles, metrics, or skills. Your profile remains the only source of facts.
No credits, no quota, no API key
There is no cloud inference call in the generation path. That means no provider key to configure, no token budget, and no external quota that can fail during an application. The trade-off is that the engine is deterministic: it assembles and emphasizes your facts rather than writing new prose from a blank page.
Where sync fits
Generation is local. Resume sync is not. When you are signed in, your master profiles and generated resumes sync to your PocketResume account so they are available on your other devices. The extension is open source, so the sync behavior is auditable.
What the engine will not do
- Invent experience, employers, titles, or metrics.
- Submit an application for you.
- Replace your judgement about which experience is relevant.
- Guarantee that a specific ATS will rank you highly.
That constraint is the point. A resume you cannot defend in an interview is not an optimization. The engine gets you to a clean, consistent draft faster; the facts stay yours.
Build the resume for the next job you open.
On-device generation, Resume Health Check, cover letters, sync, and the Job Tracker. Free until December 31, 2026.