RoleTunenever invent.Find the evidence.

RoleTune started as a personal tool. Its founder sent 164 applications and received one response. After switching to an evidence-led approach and targeting roles that actually fitted, six subsequent applications reached interview. That is one person’s search, not a promise, and the homepage says so in those words. But it explains the product’s whole attitude.

A candidate may have exactly the experience a job needs, buried three roles back, described in another industry’s vocabulary, or missing from the version of the CV they sent. They may also be applying for roles they genuinely do not fit. Rewriting everything more confidently does not tell those two situations apart, and most AI CV tools do precisely that.

The product specification set four principles: relevance, evidence, visibility and credibility. Prioritise what matters for this vacancy; trace every claim to something the candidate supplied; make relevant experience easier to see; and keep the candidate’s own voice, with no invented achievements and none of the “visionary, dynamic, results-driven” vocabulary of an inflated CV.

The identity is warm paper, ink-navy text, an indigo brand colour and an amber “tune” accent. Bricolage Grotesque for display, Geist for reading, and a tuning-dial mark that connects the name to adjusting presentation rather than inventing a new career. CSS tokens carry the palette, surfaces and status colours through the public site, the sign-in screens and the application shell alike.

The September homepage puts the product on show. A five-card strip alternates portraits with miniature product views: a fit analysis with its evidence states, and a tuned CV line labelled “Brought forward” beside the source it came from. The four How it works cards contain miniature interfaces rather than decorative icons. The strip becomes a horizontal snap-scrolling row on smaller screens. It communicates both the human context and the mechanism behind the promise.

Inside the app, a sidebar and mobile navigation connect the dashboard, Career Profile, applications, CVs, insights and billing. Score rings, verdict banners and evidence badges organise the analysis, and source references let a user inspect the basis of any recommendation. Early real use fixed real things: a slow PDF download that looked unresponsive gained a spinner and clear notifications; technical export links left the customer interface; the mobile menu became accessible; reduced-motion users get a stationary progress indicator.

A user uploads a PDF or DOCX CV. Text is extracted, then the AI layer identifies career evidence and profile details. Extraction is inspectable: a coverage report shows which sections were captured, partly captured or missed, every fact must match an excerpt from the source, and unresolved references are downgraded rather than shown as complete. The upload is not a black box; the user reviews and corrects the evidence.

Paste a job advert and the application extracts its requirements, retrieves the relevant evidence and presents a criterion-by-criterion matrix: importance, evidence strength, visibility and gap type. Gaps are categorised, from terminology and visibility problems through transferable experience to genuine gaps and hard blockers. The verdict follows explicit rules: a hard blocker means “not recommended”; otherwise a score of 60 or more means “pursue” and less means “pursue with caution”. Product rules, not validated hiring thresholds, and the point is to help someone decide whether and how to apply rather than to produce a more confident CV for every vacancy.

Recommended edits name the proposed wording and the evidence that supports it. The integrity layer resolves citations and drops suggestions without a valid source. The assembled CV is generated as a text-based A4 PDF through headless Chromium with a layout designed for straightforward parsing. ATS-readable, by design; not a guarantee of passing any particular screening system.

What is evidenced: a personal local tool evolved into a branded, multi-user application with account ownership and billing; CV recommendations connected to reviewable sources with explicit handling of uncertainty and genuine gaps; persistent workspaces with immutable revision history; reusable career evidence and private semantic retrieval; verified production PDF delivery and a specific prompt-cache reuse scenario; and a public content and SEO foundation refined from an audit of all 11 public pages.

What we do not claim: typical interview uplift, ATS bypass, guaranteed truthful output, paying customers or search growth. The first real purchase and the complete signed-in production journey for the career features are still to be confirmed. Next: those confirmations, analysis latency and cost measurement, stronger sourced editorial with real worked examples, and an organic baseline to measure against.