RESUME-SCORING-

AI-powered Resume Scoring & Job Filtering Platform. ML backend ranks resumes, scores match (0-1), and shortlists candidates automatically. Built using Gradient Boosting & TF-IDF on 9,544 resume-job pairs. REST API with Flask integrates with MERN Stack. R²: 0.59 | NDCG: 0.99 | F1: 0.86

Resume Scoring & Job Filtering Platform — API

Setup

pip install -r requirements.txt
python app.py

Server starts at: http://localhost:5000

---

Folder Structure

resume_api/
├── app.py              ← Flask API
├── requirements.txt
└── models/
    ├── model.pkl       ← Trained GradientBoosting model
    ├── scaler.pkl      ← StandardScaler
    ├── tfidf_main.pkl  ← TF-IDF for combined text
    └── tfidf_cos.pkl   ← TF-IDF for cosine similarity

---

Endpoints

1. POST /predict

Score a single resume vs job.

Request:

{
  "resume_text": "Data Engineer with 4 years experience...",
  "job_text": "Data Engineer Azure\nETL Tools\nSQL...",
  "skills": ["Python", "SQL", "Azure"],
  "skills_required": ["SQL", "ETL", "Power BI"]
}

Response:

{
  "score": 0.788,
  "percentage": "78.8%",
  "shortlist_label": "Strong Match",
  "skill_analysis": {
    "matched_skills": ["sql"],
    "missing_skills": ["etl", "power bi"],
    "match_ratio": 0.333
  }
}

---

2. POST /rank

Rank multiple resumes for one job (recruiter view).

Request:

{
  "job_text": "AI Engineer...",
  "skills_required": ["Python", "TensorFlow"],
  "resumes": [
    {"id": "r1", "resume_text": "...", "skills": ["Python", "TensorFlow"]},
    {"id": "r2", "resume_text": "...", "skills": ["JavaScript"]}
  ]
}

Response:

{
  "summary": {
    "total_resumes": 2,
    "strong_match": 1,
    "maybe": 1,
    "not_shortlisted": 0,
    "top_candidate_id": "r1"
  },
  "ranked_resumes": [
    {"rank": 1, "id": "r1", "score": 0.803, "shortlist_label": "Strong Match"},
    {"rank": 2, "id": "r2", "score": 0.674, "shortlist_label": "Maybe"}
  ]
}

---

3. POST /shortlist

Return only shortlisted candidates above a threshold.

Request:

{
  "job_text": "...",
  "skills_required": [...],
  "threshold": 0.65,
  "resumes": [...]
}

---

4. GET /health

{
  "status": "ok",
  "model": "GradientBoostingRegressor",
  "features": 513,
  "thresholds": {"strong_match": 0.75, "maybe": 0.60}
}

---

Shortlist Labels

Label	Score Range	Meaning
🟢 Strong Match	>= 0.75	Definitely interview
🟡 Maybe	0.60 - 0.75	Recruiter review
🔴 Not Shortlisted	< 0.60	Not suitable

---

Production Deployment

# Use gunicorn instead of Flask dev server
gunicorn -w 4 -b 0.0.0.0:5000 app:app

Model Performance

• R² Score: 0.59
• RMSE: 0.107
• Ranking NDCG: 0.991
• Shortlist F1: 0.863