AI Resume Parser & Scorer

I built this to solve a problem I kept running into while applying for jobs — never knowing how well my resume actually matched a job posting before hitting submit. So instead of manually comparing keywords by eye, I built a tool that does it for me using AI.

Upload a resume (PDF), paste in a job description, and it gives you a match score out of 100, points out what's missing, what's already strong, and three concrete things to fix.

Live demo: resume-scorer-nu.vercel.app

Heads up — the backend is hosted on Render's free tier, which spins down after 15 minutes of no traffic. So if nobody's used it in a while, the first request can take 30-50 seconds to wake up before it responds. After that it's fast. Just don't think it's broken if the first try feels slow.

What it does

• Drag-and-drop PDF upload
• Paste any job description
• AI compares the two and returns a structured breakdown:
  • Overall match score
  • Missing keywords from the JD
  • Existing strengths in the resume
  • 3 specific improvement tips
• Animated, responsive UI — works fine on phone or desktop
• Reset button to analyze another resume without refreshing the page

Stack

Frontend: React (Vite), Tailwind CSS v4, Framer Motion, Lucide icons, Axios

Backend: Node.js, Express, Multer for file uploads, pdf-parse for text extraction, Google Gemini API for the actual analysis

Deployment

• Frontend deployed on Vercel
• Backend deployed on Render

Running it locally

You'll need Node 18+ and a free Gemini API key from aistudio.google.com/app/apikey (takes about a minute, just needs a Google login, no card).

Clone it:

git clone https://github.com/Mohammed-Omer-S/resume-scorer.git
cd resume-scorer

Backend:

cd backend
npm install
cp .env.example .env

Open .env and drop your key in:

PORT=5000
GEMINI_API_KEY=your_key_here

Then:

npm run dev

Frontend, in a separate terminal:

cd frontend
npm install

Create a .env file inside frontend with:

VITE_API_URL=http://localhost:5000

Then:

npm run dev

Open http://localhost:5173 and try it.

How it actually works

1. PDF gets uploaded, backend extracts the raw text with pdf-parse
2. That text + the job description get sent to Gemini with a strict prompt telling it to return only JSON in a specific shape
3. Backend parses that JSON and sends it to the frontend
4. Frontend renders it as score bars, badges, and a tips list

Nothing gets saved anywhere — no database, no accounts. Each analysis is a one-off, which was a deliberate choice to keep this simple and stateless.

Folder structure

resume-scorer/
backend/
  server.js
  routes/
    analyze.js
  utils/
    aiPrompt.js
  .env.example
frontend/
  src/
    components/
    App.jsx
    index.css

License

MIT — see LICENSE file.

---

Built by Mohammed Omer. If you spot a bug or have a suggestion, feel free to open an issue.