AI Resume Analyzer

A production-grade full-stack AI SaaS application that analyzes resumes against job descriptions using Groq LLM, Google Gemini embeddings, and ATS-style scoring algorithms.

Portfolio-ready project: Demonstrates AI integration, full-stack architecture, JWT auth, Redis caching, vector similarity, and scalable system design.

---

🏗️ Architecture

┌─────────────────────────────────────────────────────────┐
│                     FRONTEND (React)                    │
│  Landing → Auth → Dashboard → Analyze → Results         │
│  Deployed on: Vercel                                    │
└────────────────────────┬────────────────────────────────┘
                         │ HTTPS / REST
┌────────────────────────▼────────────────────────────────┐
│                  BACKEND (Node/Express)                 │
│                                                         │
│  /auth     → JWT auth (login/register)                  │
│  /resume   → Upload + parse PDF                         │
│  /analysis → AI pipeline (Groq + Gemini)                │
│  /user     → Profile + usage tracking                   │
│                                                         │
│  Deployed on: Render                                    │
└──────────┬──────────────────────┬───────────────────────┘
           │                      │
┌──────────▼──────┐    ┌──────────▼──────────────────────┐
│   MongoDB Atlas │    │      AI Layer                   │
│                 │    │                                │
│  Users          │    │  Groq (llama-3.3-70b)           │
│  Resumes        │    │   → Resume analysis (ATS)        │
│  Analyses       │    │   → Resume rewriting             │
└─────────────────┘    │                                │
           │           │  Gemini (embedding-001)         │
┌──────────▼──────┐    │   → Semantic similarity         │
│  Redis (Upstash)│    └────────────────────────────────┘
│  Cache (1hr TTL)│
└─────────────────┘

---

🚀 Features

Feature	Details
ATS Score (0–100)	Groq LLM evaluates resume vs JD
Semantic Similarity	Gemini embeddings + cosine similarity
Section-wise Scores	Skills, Experience, Education, Projects
Composite Score	50% LLM + 30% semantic + 20% sections
Missing Skills Detection	Extracted directly from JD
Actionable Suggestions	Specific improvements (not generic)
ATS Tips	Resume optimization guidance
AI Resume Rewriter	Groq rewrites resume for better matching
JWT Auth	Secure login + refresh tokens
Redis Caching	Same resume+JD → instant result
Rate Limiting	Prevents abuse

---

🧠 AI Pipeline

Resume + Job Description
        │
        ▼
Groq → LLM Analysis (ATS Score, Skills, Suggestions)
        │
        ▼
Groq → Section-wise scoring
        │
        ▼
Gemini → Generate embeddings (resume + JD)
        │
        ▼
Cosine Similarity → Semantic Score
        │
        ▼
Weighted Average:
50% LLM + 30% Semantic + 20% Sections
        │
        ▼
Final Composite Score

---

🛠️ Tech Stack

Frontend

• React + Vite
• Tailwind CSS
• Zustand
• TanStack Query

Backend

• Node.js + Express
• MongoDB + Mongoose
• Redis (Upstash)
• Groq SDK
• Google Generative AI (Gemini)
• JWT Auth + bcrypt
• Multer + pdf-parse

---

📁 Project Structure

ai-resume-analyzer/
├── client/        # React frontend
└── server/        # Express backend
    ├── routes/
    ├── controllers/
    ├── services/
    │   └── aiService.js   # Groq + Gemini logic
    ├── models/
    └── middleware/

---

⚙️ Environment Variables

Server (server/.env)

NODE_ENV=production
PORT=5000

MONGODB_URI=your_mongodb_uri

JWT_SECRET=your_secret
JWT_REFRESH_SECRET=your_secret

GROQ_API_KEY=your_groq_key
GEMINI_API_KEY=your_gemini_key

REDIS_URL=your_redis_url
CLIENT_URL=http://localhost:3000

---

Client (client/.env)

VITE_API_URL=http://localhost:5000/api/v1

---

▶️ Local Setup

git clone https://github.com/yourusername/ai-resume-analyzer.git
cd ai-resume-analyzer

npm run install:all
npm run dev

---

📊 API Example

Analyze Resume

POST /api/v1/analysis/analyze

{
  "resumeId": "123",
  "jobDescription": "We are looking for a Software Engineer..."
}

---

Response

{
  "compositeScore": 73,
  "semanticScore": 68,
  "llmAnalysis": {
    "score": 76,
    "missing_skills": ["Docker", "Kubernetes"],
    "suggestions": ["Add quantified achievements"]
  }
}

---

💡 Why Groq + Gemini?

• Groq → ultra-fast, low-cost LLM for analysis & rewriting
• Gemini → high-quality embeddings for semantic matching

👉 This hybrid approach balances:

• ⚡ Speed
• 💰 Cost
• 🧠 Accuracy

---

📈 Limits (Free Tier)

Service	Limit
Groq	~14K requests/day
Gemini	~1500 embeddings/day

👉 ~700–750 resume analyses/day (without caching)

---

🚀 Deployment

Backend → Render

• Build: npm install
• Start: npm start

Frontend → Vercel

• Build: npm run build

Database → MongoDB Atlas

Cache → Upstash Redis

---

🧠 System Design Highlights

• Hybrid AI architecture (LLM + embeddings)
• Caching layer (Redis)
• Stateless backend (scalable)
• Composite scoring (robust evaluation)

---

🎯 Interview Explanation

“I used Groq for fast and cost-efficient LLM tasks like resume analysis and rewriting, and Gemini for embeddings to compute semantic similarity. This hybrid design improves performance while maintaining accuracy.”

---

📌 Future Improvements

• Queue system (BullMQ)
• WebSocket updates
• Resume templates export
• Payment integration

---

📜 License

MIT