Shortlisted - Built to get you shortlisted.

A resume keyword analyzer and job application tracker. Paste your resume and a job description — get a match score, keyword gaps, filler word detection, and actionable fixes in under a second. Built for job seekers who are tired of getting ghosted.

🔗 Live demo: shortlisted-beta.vercel.app

---

What it does

Resume Analyzer (free, unlimited)

• Compares your resume against a job description using TF-IDF keyword scoring
• Shows which keywords are missing, grouped by category (Tools, Technical, Soft Skills)
• Detects 60+ filler phrases ("responsible for", "team player", "hardworking") with specific rewrite suggestions
• Weighted match score — technical skills and tools rank higher than soft skills
• Runs entirely client-side — your resume text never leaves your browser

Application Tracker

• Kanban board: Applied → Screening → Interview → Offer → Rejected
• Drag-and-drop status updates with optimistic UI
• Notes, JD URL, and applied date per application
• Follow-up reminders and analytics (in progress)

---

Architecture

Analysis engine (src/lib/engine/)

The core of the product. Pure TypeScript, zero dependencies, runs in the browser.

tokeniser.ts      → lowercase, strip punctuation, expand aliases (k8s → kubernetes)
stopwords.ts      → 400+ word filter including resume noise ("responsible", "experience")
stemmer.ts        → lightweight Porter Stemmer, NOSTEM set protects technical terms
tfidf.ts          → TF-IDF scoring against pre-computed reference IDF corpus
gapAnalyser.ts    → compares top 30 JD keywords vs resume tokens
fillerDetector.ts → 60-phrase blacklist with replacement suggestions
scorer.ts         → weighted match score (technical=1.0, tool=1.0, general=0.5, soft=0.3)
index.ts          → public API: analyseResume(resumeText, jdText) → AnalysisResult

Key decisions:

• No AI in MVP — deterministic logic is fast, free, and debuggable. Claude Haiku planned for v2 bullet rewrites.
• Signal-based filtering — unknown terms only pass if they appear 2+ times in the JD AND exceed a TF-IDF threshold. This kills location names, company names, and posting metadata without a blocklist.
• Category weighting — a JD mentioning "kubernetes" once matters more than "communication" three times. Weights are applied in the scorer, not the extractor.

Stack

Layer	Choice	Why
Framework	Next.js 16 (App Router)	One codebase for frontend + API routes. Vercel deployment is zero-config.
Styling	Tailwind CSS	Utility-first keeps styling colocated. No dead CSS.
Auth	Supabase Auth	Google OAuth in 3 clicks. Auth UID = database UID — no glue code. RLS enforced at DB level.
Database	Supabase Postgres	Full SQL with joins and aggregations. RLS means the DB enforces access control, not the app.
Hosting	Vercel (Hobby)	Free tier, auto-deploys on push, global CDN, preview URLs per PR.
Emails	Resend	Deferred — needs custom domain first.

Why the analyzer runs client-side

Most resume tools send your resume to a server. That means your employment history, contact details, and career gaps are stored somewhere you don't control.

Running analysis in the browser means:

• Zero server cost per analysis (important at MVP stage with no revenue)
• No privacy risk — we never see the content
• Instant results — no network round-trip
• Scales to any number of free users for free

The tradeoff is that we can't do server-side ML or store scan history for free users. That's an acceptable tradeoff at this stage.

---

Running locally

Prerequisites: Node.js 22, pnpm

git clone https://github.com/RogueStar03/shortlisted
cd shortlisted
pnpm install

Create .env.local:

NEXT_PUBLIC_SUPABASE_URL=your_supabase_url
NEXT_PUBLIC_SUPABASE_PUBLISHABLE_KEY=your_publishable_key
SUPABASE_SECRET_KEY=your_secret_key
NEXT_PUBLIC_APP_URL=http://localhost:3000

Set up the database — run these SQL blocks in your Supabase SQL editor in order:

1. supabase/migrations/001_profiles.sql
2. supabase/migrations/002_applications.sql
3. supabase/migrations/003_application_events.sql
4. supabase/migrations/004_reminders.sql

pnpm dev

App runs at http://localhost:3000.

Run the engine smoke test:

pnpm tsx src/lib/engine/engine.test.ts

---

Project structure

src/
├── app/
│   ├── page.tsx                    # Landing page
│   ├── auth/                       # Sign in / sign up + OAuth callback
│   ├── analyze/                    # Resume + JD input (client-side analysis)
│   ├── results/                    # Analysis results UI
│   ├── tracker/                    # Kanban board (Placement Pack)
│   ├── privacy/ terms/ contact/    # Legal pages
│   └── api/applications/           # REST API for tracker CRUD
├── components/
│   ├── ui/                         # Button, Card, Badge, ScoreBar
│   ├── layout/                     # Navbar
│   └── landing/                    # HowItWorks animated section
└── lib/
    ├── engine/                     # Analysis engine (8 files, zero dependencies)
    ├── supabase/                   # Browser + server clients, applications queries
    └── constants/                  # Color tokens

---

Status

Feature	Status
Resume analyzer	✅ Complete
Auth (Google + email)	✅ Complete
Landing page	✅ Complete
Application tracker (Kanban)	🔄 In progress
Follow-up reminders	⏳ Planned
Analytics dashboard	⏳ Planned
PDF resume upload	⏳ v2
AI bullet rewrites (Claude Haiku)	⏳ v2

---

Built with

This project was built collaboratively with Claude (Anthropic) over several sessions — architecture decisions, engine design, UI components, and all code written together. The analysis engine, stack choices, and product decisions are documented in detail across the build sessions.

---

License

MIT — use it, learn from it, build on it.