📄 AI-Powered Resume Screener & Ranker

An automated hiring tool that uses Natural Language Processing (NLP) and Machine Learning to screen, score, and rank candidate resumes against a job description.

Build By:

• Priyanshu Kumawat - 22BCON1393
• Rohit Kumar Saini - 22BCON1431
• Hemant Singh - 22BCON1402
• Manjeet Jahkar - 22BCON1454

---

🚀 Key Features

• Semantic Analysis: Uses sentence-transformers to compute the Cosine Similarity between the Job Description and Resume, detecting relevant candidates even if they use different terminology.

• Keyword Verification: Checks for mandatory hard skills (e.g., Python, SQL) using Regex to ensure technical compliance.

• Experience Parsing: Automatically extracts years of experience from text patterns to weight seniority.

• Ranked Output: Generates a clean CSV report (ranking_report.csv) sorting candidates from highest to lowest fit.

  Candidate	Final Score	AI Score	Keyword Match	Experience Score	Missing Skills
  Ashish Ohri.pdf	87.94	39.7	100.0	100.0	None
  Juan Josecarin.pdf	77.83	39.2	66.7	100.0	Pandas, NLP
  Saran Wong.pdf	65.76	28.8	33.3	100.0	Machine Learning, Scikit-Learn, Pandas, NLP
  Bhavesh Wadhwani.pdf	56.77	33.9	83.3	50.0	Machine Learning
  Yunlong Jiao.pdf	37.55	37.7	100.0	0.0	None
  Zain Khalid.pdf	36.67	33.3	16.7	50.0	Python, Machine Learning, Scikit-Learn, Pandas, NLP
  Sunmarg_resume.pdf	28.55	17.8	0.0	50.0	Python, Machine Learning, Scikit-Learn, Pandas, NLP, SQL
  Deep Mehta.pdf	20.88	29.4	50.0	0.0	Scikit-Learn, Pandas, NLP
  FLorne.pdf	8.58	17.9	16.7	0.0	Python, Machine Learning, Scikit-Learn, Pandas, NLP
  Kartik tomer.pdf	4.11	20.6	0.0	0.0	Python, Machine Learning, Scikit-Learn, Pandas, NLP, SQL

🛠️ Tech Stack

• Language: Python 3.x
• NLP Model: all-MiniLM-L6-v2 (via Sentence-Transformers)
• PDF Parsing: pypdf
• Data Handling: pandas
• Pattern Matching: re (Regular Expressions)

⚙️ How It Works (The Logic)

The system calculates a Total Score (0-100%) based on a weighted average of three metrics:

1. Semantic Score (20%): The text is converted into high-dimensional vectors (Embeddings). We calculate the Cosine Similarity between the Job Description vector and the Resume vector.
2. Keyword Match Score (30%): Checks for the presence of required skills defined in the configuration.
3. Experience Score (50%): Extracts "Years of Experience" (e.g., "5+ years") and normalizes it against the job requirement.

📂 Project Structure

Resume-Screener/
├── models/                  # Contains SBERT model files
├── resumes/                 # Folder to store candidate PDF files
│   ├── candidate_1.pdf
│   └── candidate_2.pdf
├── main.py                  # Main script containing the logic
├── requirements.txt         # List of dependencies
├── ranking_report.csv       # Output file (Generated after running)
├── LICENSE                  # MIT License
└── README.md                # Documentation