Skip to content

1. Overview

← Home · Next: Installation →

This page explains, in plain language, what Maestro AI does and how its parts work together. If you just want to install it, you can skim this and jump to the Installation Guide — but five minutes here will make everything else clearer.

What problem does it solve?

Applying for jobs is repetitive and slow: find the posting, decide whether it's worth your time, rewrite your resume for it, write a cover letter, and keep track of where you've applied. Maestro automates the tedious parts and leaves the decisions to you.

It does five things:

  1. Discovers open jobs from company career pages and job APIs.
  2. Scores each job for how well it fits you — twice, in fact: once against the role you're targeting and once against your actual background.
  3. Builds a tailored resume and cover letter for the jobs you choose, using AI agents that draft, critique, refine, and fact-check the result.
  4. Lets you review and refine everything through a web dashboard.
  5. Tracks every application's status, from "not applied" through "offer" or "rejected."

The two halves

Maestro is made of two cooperating systems that share a single database.

flowchart TD
    DB[("Google Sheets database<br/>+ Google Drive files")]
    BE["Backend pipeline<br/>(n8n + AI agents)"]
    DASH["Maestro dashboard<br/>(Next.js web app)"]
    SCHED["Scheduler<br/>(Node worker)"]
    DB <--> BE
    DB <--> DASH
    BE <-->|webhooks| DASH
    SCHED -->|fires on a schedule| BE

Backend pipeline (n8n)

n8n is a visual automation tool. Maestro's backend is a set of n8n workflows — think of each as a flowchart that runs automatically. The pipeline contains ten AI agents, each a specialist:

Agent Job
1 — Resume Builder Drafts a tailored resume from your master profile + the job description
2 — Cover Letter Builder Drafts a matching cover letter
3 — Resume Verifier Fact-checks the resume against your profile (no invented claims)
4 — Cover Letter Verifier Fact-checks the cover letter
5 — Critic Critiques the resume like a recruiter would
6 — Resume Refiner Rewrites the resume to address the critique
7 — Resume Scorer Scores the final resume's fit (0–100)
8 — Discovery Scorer Scores newly discovered jobs for fit
9 — Resume Fine-Refiner Applies your manual refinement instructions on later passes
10 — Job Ranker A fast first-pass ranker that keeps the discovery funnel open

These agents don't talk to AI providers directly. They all route through one shared module, Call LLM, which supports Anthropic Claude, OpenAI, and Google Gemini and handles errors, retries, and cost tracking in one place.

Maestro dashboard (Next.js)

The dashboard is the human side. It reads the same Google Sheet the pipeline writes to, and it sends instructions back to the pipeline through webhooks (a webhook is just a URL the dashboard calls to say "please run this"). From the dashboard you can:

  • Browse discovered jobs and their fit scores
  • Select jobs and click Build to generate applications
  • Read the generated resume, see the critic's notes and verifier's checks, and request refinements
  • Save a polished .docx resume to Google Drive
  • Track every application's status

The scheduler

n8n can't read a schedule out of your spreadsheet on its own, so Maestro includes a tiny standalone scheduler — a Node program in its own container. It reads how often you want to discover jobs (a setting in your config) and fires the discovery pipeline on that cadence. You can also trigger discovery manually from the dashboard at any time.

How the data flows

Everything is coordinated through one Google Sheets file with many tabs (the full list is in the Database Reference). The most important tabs:

  • config — all your settings (your name, target roles, which AI models to use, how often to run discovery).
  • jobs — every job discovered or submitted, plus its scores and status.
  • resumes — every resume version generated, including refinements.
  • agent_outputs — the raw critic/verifier/scorer output behind each resume.
  • model_usage — a per-call log of tokens and cost, so you always know what you're spending.

The pipeline writes; the dashboard reads and writes back; nothing is hidden.

Two scoring axes (why fit scores come in pairs)

A subtle but important design point: every discovered job gets two fit scores.

  • Target fit — does this job match the role you're searching for?
  • Background fit — does this job match the experience you actually have?

These can legitimately differ. If you're an experienced executive applying to a more junior IC role on purpose, the target fit might be high while the background fit flags the seniority gap. Keeping them separate stops one from masking the other.

What's automated vs. what you decide

Maestro does automatically You decide
Find and score jobs Which jobs to build applications for
Draft, critique, refine, verify resumes Whether a resume is good enough to send
Track cost and tokens Which AI models and budget to use
Generate .docx files When and where to actually apply

Nothing is sent anywhere on your behalf. Maestro prepares; you act.

← Home · Next: Installation →