CV Analyzer

A three-agent pipeline behind one upload.

The frontend is a clean three-step flow — upload, results, optimize — but the interesting work happens in an n8n pipeline of three cooperating AI agents. Deliberately, it asks for the job description as text rather than a link: scraping job posts hurts input quality, so the design trades convenience for a cleaner signal.

Analyzer, persona, matcher.

Each agent has one job, and the handoffs between them are the design:

• A CV Analyzer uses vision to extract structured text from the uploaded PDF; an HR Persona builds an ideal-candidate profile from the job description.
• A Matcher compares the two with structured JSON output — its system prompt alone runs 320 lines — producing the score, categorized feedback, and dynamic follow-up questions.
• The frontend builds its forms from those questions at runtime with generated schemas; a fallback node gracefully converts option-less questions into free text, so it never breaks on an unexpected job.
• An optimization pass adds two more agents that generate the final CV as both PDF and DOCX.