An AI-assisted placement preparation platform: mock interviews with per-answer grading, resume ↔ job-description analysis, a community question bank, an application tracker, and company research.

Built with Next.js 16 (App Router), TypeScript, MongoDB/Mongoose, Auth.js v5, Tailwind v4 + shadcn-style components, and Groq (Llama 3.1) for every AI call. An optional FastAPI service scrapes public interview experiences.

---

Quick start

npm install
cp .env.example .env.local     # fill in MONGODB_URI, AUTH_SECRET, GROQ_API_KEY
npm run dev                    # http://localhost:3000

Command	What it does
npm run dev	Development server
npm run build / npm start	Production build and server
npm run lint	ESLint
npm test	Unit tests (node --test, no database needed)

Required environment variables

Variable	Why
MONGODB_URI	Mongo connection string (DATABASE_URL also accepted)
AUTH_SECRET	Signs the session JWT — openssl rand -base64 32
GROQ_API_KEY	Resume analysis, interview generation and grading

Optional: GOOGLE_CLIENT_ID / GOOGLE_CLIENT_SECRET, GITHUB_ID / GITHUB_SECRET (OAuth providers are only registered when their pair is present), GROQ_MODEL, PYTHON_SERVICE_URL.

---

Layout

app/
  (auth)/login, (auth)/register     Auth screens
  api/                              Route handlers (the entire backend)
  dashboard/                        Product screens
components/
  brand/Logo.tsx                    The logo — mark, wordmark and lockup (single source of truth)
  ...                               applications, questions, resume, landing, ui primitives
lib/                                auth, db, groq, http/api helpers, quota, resume parsing, validation
models/                             Mongoose schemas
services/scraping-service/          Optional FastAPI research service
tests/                              Unit tests for the pure logic

LEARNING.md is the full technical walkthrough (every flow, every design decision). INTERVIEW.md prepares you to defend the project in an interview.

---

Brand

The mark is a P whose stem doubles as a rising arrow — preparation turning into progress — set in a rounded ink badge with the letter's counter picked out in crimson. It is drawn on a 40×40 grid from three paths, so it stays crisp from a 16px favicon to a hero lockup.

Token	Value	Use
Ink	#141210	Badge, glyph on light surfaces
Paper	#FFFFFF	Glyph on ink, badge on dark surfaces
Crimson	#E8174A	The counter, and AI in the wordmark

Import from components/brand/Logo.tsx — never hand-roll a mark:

import Logo, { LogoMark, Wordmark } from '@/components/brand/Logo';

<Logo />                                   // lockup, ink badge (default)
<Logo size="lg" variant="paper" />         // white badge + hard ink rule
<Logo variant="onDark" />                  // for the sidebar and ink sections
<LogoMark variant="mono" className="h-5 w-5" />   // inherits currentColor

Static assets live in public/: logo-mark.svg, logo-mark-paper.svg, logo-mark-mono.svg, logo.svg (lockup), logo.png, logo-lockup.png and og-image.png. Favicons are served from the app dir (app/favicon.ico, app/icon.svg, app/apple-icon.png).

Every asset is regenerated from one geometry definition in scripts/build-brand-assets.py — edit the mark there, never by hand:

python3 scripts/build-brand-assets.py

---

Optional research service

cd services/scraping-service
pip install -r requirements.txt
uvicorn main:app --reload --port 8000     # then set PYTHON_SERVICE_URL=http://localhost:8000

Without it, company research degrades to curated source links instead of failing.

---

Deployment

Deploy the Next.js app anywhere that runs Node (Vercel being the obvious choice): set the environment variables above, point AUTH_URL at the deployed origin, and add the deployed URL to ALLOWED_ORIGINS on the Python service if you run one.