Job Portal

A two-sided job board: job seekers browse and apply to listings, recruiters sign in to post openings and review applicants. Built as a React client plus an Express/MongoDB API, with Clerk handling authentication for both sides.

🔗 Live demo: https://job-portal-client-alpha-sand.vercel.app API: https://job-portal-server-opal-seven.vercel.app

Current state

The frontend is fully built out — job listing/search, an apply flow with resume upload, a recruiter dashboard for posting and managing jobs and reviewing applicants. The backend so far only implements the Clerk user-sync webhook (creating/updating/deleting a User document when someone signs up via Clerk) — the job/application CRUD endpoints the frontend UI expects aren't wired up yet, so the job listings you see on the live demo are static sample data (client/src/assets/assets.js), not served from the database. That's the natural next thing to build on top of this.

Auth and the Apply flow are real, not mocked: Recruiter Login and the general Login button both go through actual Clerk sign-in/sign-up (there's one shared account system - "recruiter" isn't a separate credential store, it's just the same signed-in user visiting /dashboard, which is now guarded and redirects home if you're not signed in). "Apply Now" checks you're signed in (prompting sign-in if not) and confirms the submission with a toast - but since there's no Application model on the backend yet, that confirmation isn't persisted anywhere. Both of these were previously inert UI with no logic behind the buttons at all.

Tech stack

Layer	Technology
Frontend	React 19, Vite, Tailwind CSS, React Router, Quill (rich text), Clerk (auth)
Backend	Node.js, Express, Mongoose/MongoDB, Clerk webhooks (via Svix), Sentry (error tracking)
Hosting	Vercel (client and server as separate projects)
CI/CD	GitHub Actions

Security note

This repo's .env files were committed to git with a live, credentialed MongoDB connection string in it — a real exposure, since public GitHub repos get scraped for exactly this pattern within minutes. Both .env files have been removed from git tracking (.env.example files added in their place, .gitignore updated) but the MongoDB Atlas password should still be rotated (Database Access → the user → Edit → Password) — untracking a file going forward doesn't undo something that was already public, and there's no way to know if it was scraped before this fix.

Running it locally

Prerequisites: Node.js ≥ 18, a MongoDB Atlas cluster (or local MongoDB), a Clerk application (for both frontend and backend keys).

git clone https://github.com/mkhalidh/Job_Portal.git
cd Job_Portal

Backend:

cd server
npm install
cp .env.example .env   # fill in MONGODB_URI and CLERK_WEBHOOK_SECRET
npm run server          # nodemon, http://localhost:5000

Frontend (separate terminal):

cd client
npm install
cp .env.example .env   # fill in VITE_CLERK_PUBLISHABLE_KEY
npm run dev              # http://localhost:5173

Deployment

Client and server are deployed as two independent Vercel projects (the server can't be a "function" inside the client's project the way a Next.js API route would be — it's a standalone Express app deployed via the legacy builds/routes format in server/vercel.json, since it's a plain app.listen() server rather than an exported request handler).

MongoDB Atlas note: Vercel's serverless functions don't have a static outbound IP on the free tier, so Atlas's IP allowlist needs 0.0.0.0/0 (Allow Access from Anywhere) added under Network Access, or every database connection from the deployed server fails.

CI/CD runs via GitHub Actions:

• .github/workflows/ci.yml — lints and builds the client, and does an install + syntax check on the server, on every push/PR
• .github/workflows/deploy.yml — deploys both client and server to Vercel on every push to main. Needs six repo secrets: VERCEL_CLIENT_TOKEN, VERCEL_CLIENT_ORG_ID, VERCEL_CLIENT_PROJECT_ID, VERCEL_SERVER_TOKEN, VERCEL_SERVER_ORG_ID, VERCEL_SERVER_PROJECT_ID (the client and server are separate Vercel projects, so separate tokens/IDs — a single Vercel token works for both if you'd rather reuse one).

A note on Sentry

server/config/instrument.cjs originally included @sentry/profiling-node for CPU profiling. That package ships a platform-specific native binary that Vercel's serverless bundler didn't include in the deployed function, which crashed every single request with Cannot find module './sentry_cpu_profiler-...'. Profiling was removed entirely (basic Sentry error tracking doesn't need it, and native-binary profiling doesn't fit ephemeral serverless functions well regardless) - if you re-add it, budget time to solve the bundling problem first.

Project structure

Job_Portal/
├── client/
│   ├── src/
│   │   ├── pages/           # Home, ApplyJob, Application, AddJobs,
│   │   │                     Dashboard, ManageJobs, ViewApplications
│   │   ├── components/      # Navbar, Hero, JobListing, JobCard, etc.
│   │   └── context/          # AppContext - jobs list, search, auth state
│   └── vercel.json           # SPA rewrite for React Router
├── server/
│   ├── controller/webhook.js  # Clerk user.created/updated/deleted sync
│   ├── models/User.js
│   ├── config/db.js            # Mongoose connection
│   └── vercel.json              # legacy builds/routes config
└── .github/workflows/           # CI/CD

License

ISC