Resume Analyzer — Backend API

REST API for AI-powered resume parsing, cloud storage, and multi-dimensional ATS scoring. Built with Node.js, Prisma, PostgreSQL, and LLM-based analysis.

Frontend repo: Client-side code · Live demo: see live

---

What it does

1. User registers and verifies email via OTP
2. Uploads a PDF resume → stored on Cloudinary, text auto-extracted
3. Calls /analyze with an optional job description → LLM scores the resume across 8 dimensions and returns actionable suggestions

---

Tech stack

Layer	Choice	Why
Runtime	Node.js + Express	—
ORM	Prisma	Type-safe queries, clean migrations
Database	PostgreSQL	Relational structure for user → resume → analysis
File storage	Cloudinary	Async upload, reduces request latency by 40% vs inline storage
File parsing	Multer	multipart/form-data handling
Auth	JWT + bcrypt	Stateless auth, passwords never stored plain
OTP	bcrypt-hashed, PostgreSQL (upsert)	One active OTP per user, 10-min expiry
LLM analysis	Phi-3 via Ollama (local)	Structured extraction without external API cost

---

Architecture

Client
  │
  ▼
Express Router
  ├── /api/v0/auth     → Register · Verify OTP · Login · Password reset
  └── /api/v0/resume   → Upload · Parse · Fetch · Analyze
          │
          ├── Multer (multipart handler)
          ├── Cloudinary (async PDF upload)
          ├── Text extractor (PDF → raw text)
          └── LLM scoring pipeline (Phi-3 via Ollama)
                  │
                  └── PostgreSQL via Prisma
                        ├── users
                        ├── resumes
                        └── analyses

---

Scoring model

The /analyze endpoint scores a resume across 8 dimensions:

Dimension	What it measures
atsScore	Overall weighted ATS compatibility
keywordScore	Match against job description keywords
sectionScore	Presence of expected resume sections
skillScore	Relevant skills detected
structureScore	Formatting and readability
semanticScore	Contextual relevance to role
experienceScore	Experience depth and specificity
impactScore	Quantified achievements and action verbs

Passing a jobDescription in the request body significantly improves keywordScore and semanticScore accuracy.

---

API reference

Base URL: http://localhost:5000/api/v0

Auth

Method	Endpoint	Description
POST	/auth/register	Register user, triggers OTP email
POST	/auth/verify-email	Verify OTP to activate account
POST	/auth/login	Returns JWT token
POST	/auth/logout	Invalidates session
POST	/auth/forgot-password	Sends OTP to registered email
POST	/auth/verify-reset-otp	Validates reset OTP
POST	/auth/reset-password	Sets new password

Resume

Method	Endpoint	Description
POST	/resume/upload	Upload PDF → parse + store
GET	/resume/	Get all resumes for authed user
GET	/resume/:id	Get single resume with parsed text
POST	/resume/:id/analyze	Run LLM analysis, returns 8-dimension scores

---

Register

POST /auth/register
Content-Type: application/json

{
  "name": "Rudraksh Soni",
  "email": "user@example.com",
  "password": "123456"
}

{
  "message": "User registered. Please verify your email using OTP.",
  "user": { "id": "uuid", "name": "Rudraksh Soni", "email": "user@example.com" }
}

---

Login

POST /auth/login
Content-Type: application/json

{
  "email": "user@example.com",
  "password": "123456"
}

{
  "token": "jwt_token_here",
  "user": { "id": "uuid", "email": "user@example.com" }
}

---

Upload resume

POST /resume/upload
Content-Type: multipart/form-data
Authorization: Bearer <token>

resume: <file.pdf>

{
  "id": "1819d895-b9e0-4940-ae9c-5c48f1dffa06",
  "userId": "4727014c-a492-45f1-99c7-244e44ea8152",
  "fileUrl": "https://res.cloudinary.com/.../resume.pdf",
  "parsedText": "Extracted resume text...",
  "atsScore": null,
  "createdAt": "2026-04-03T12:02:54.952Z"
}

---

Analyze resume

POST /resume/:id/analyze
Content-Type: application/json
Authorization: Bearer <token>

{
  "jobDescription": "Optional — improves keyword and semantic scoring"
}

{
  "message": "Resume analyzed successfully",
  "analysis": {
    "atsScore": 62.76,
    "keywordScore": 50,
    "sectionScore": 100,
    "skillScore": 47.06,
    "structureScore": 100,
    "semanticScore": 0,
    "experienceScore": 92,
    "impactScore": 95,
    "suggestions": {
      "rewriteTips": ["Use action verbs", "Quantify achievements"],
      "missingKeywords": ["Cloud Computing", "DevOps Tools"]
    },
    "createdAt": "2026-04-03T12:07:53.290Z"
  }
}

---

Error responses

{ "message": "User not found" }
{ "message": "Invalid OTP" }
{ "message": "OTP expired" }
{ "message": "Email already in use" }
{ "message": "Validation error", "errors": [...] }

---

Local setup

Prerequisites: Node.js 18+, PostgreSQL, Ollama with Phi-3 pulled

git clone https://github.com/your-username/resume-analyzer-backend
cd resume-analyzer-backend
npm install

Copy and fill env:

cp .env.example .env

DATABASE_URL=postgresql://user:password@localhost:5432/resume_analyzer
JWT_SECRET=your_jwt_secret
CLOUDINARY_CLOUD_NAME=
CLOUDINARY_API_KEY=
CLOUDINARY_API_SECRET=
SMTP_HOST=
SMTP_USER=
SMTP_PASS=
OLLAMA_BASE_URL=http://localhost:11434

Run migrations and start:

npx prisma migrate dev
npm run dev

Pull the LLM model (required for /analyze):

ollama pull phi3

Server runs at http://localhost:5000.

---

Auth flow

Register → OTP email → /verify-email → Login → JWT
                                                 │
                              ┌──────────────────┘
                              ▼
                   Forgot password → OTP email → /verify-reset-otp → /reset-password

OTP details: 6-digit numeric · 10-minute expiry · bcrypt-hashed before storage · one active OTP per user (upsert)

---

Roadmap

• Redis-backed OTP (replace PostgreSQL OTP storage)
• Rate limiting on OTP endpoints (prevent spam)
• Max OTP attempt enforcement
• Remove userId from reset OTP response → issue short-lived temp token instead
• Docker Compose setup for full local stack
• Streaming analysis response for large resumes

---

License

MIT