From paste to PDF in five minutes

Paste any JD — a URL or raw text — and Stet parses it before you upload anything else. Within a couple of seconds you see the parsed signals: must-have skills, role title, seniority, and a confidence score on each. The parser uses a typed BAML schema, so the structure is consistent across job boards.

Why parse first? Because the JD shapes everything that follows. Knowing the must-have skills before you upload lets the agent tailor straight to the JD on the first pass rather than retroactively guessing.

Drop a PDF or DOCX. Stet extracts a structured tree — sections, bullets, dates, companies — using a forked Reactive-Resume parser. You don’t rebuild your résumé in a form; the existing one is the source of truth.

Your file is stored in Supabase under a row-level-security policy scoped to your account. See the security page for retention windows and the no-training contract that covers the inference path.

This is the core moment. The workspace splits into Studio (chat, ~30%) and Vellum (the résumé canvas, ~70%). As you talk to the agent — “tailor my experience for this backend role” — it edits the canvas section by section. A 2px ember outline glows around the section being mutated so you know exactly what changed.

You drive the chat. The agent doesn’t ghost-write — it tailors what’s already there. It will not invent job titles, dates, or experience. If you ask it to fabricate, it refuses. The constraint is opinionated, but it’s also why the output passes ATS scrutiny instead of triggering keyword-stuffing flags.

Every change is reversible. The chat history persists, so you can resume mid-tailor on another device.

When the variant is ready, Stet computes a three-stage ATS scorecard: keywords (do the JD’s must-have skills appear, and where), formatting (no images of text, no multi-column tricks, no parsing-hostile fonts), and experience match (does the rewritten experience map cleanly to the role level the JD targets).

Each sub-score has a short explanation, so the number is auditable instead of a black box. Export an ATS-clean PDF and you’re done. The same PDF that looks right in your browser is what the recruiter — and their ATS — will see.