Resume Screener

A resume parsing and scoring system built with Streamlit and LLM technology. The system extracts structured information from resumes and provides a comprehensive scoring based on multiple components.

Features

1. Resume Parsing

• Main Information: Uses LLM to extract structured data including:

  • Personal Information (name, contact details, etc.)
  • Academic Performance
  • Technical Skills
  • Projects and Internships

• Extra-Curricular Activities: Uses pattern-based extraction for:

  • Leadership Roles
  • Awards and Achievements
  • Certifications
  • Activities and Involvement
  • Language Proficiency

2. Scoring System

The system employs a 100-point scoring mechanism divided into four components:

A. Academic Performance (20%)

• Formula: (CGPA × 0.75) + ((10 - Std Dev) × 0.25)
• Considers both CGPA and consistency across semesters
• Standard deviation of SGPAs used to measure consistency
• Score normalized to maximum of 20%

B. Technical Skills (35%)

Points allocated based on number of skills in each category:

• 0 skills: 0 points
• 1-3 skills: 1 point
• 4+ skills: 2 points

Categories:

• Programming Languages
• Frameworks
• Databases
• Other Technologies
• Knowledge Areas

Maximum raw points: 10 (2 points × 5 categories) Final score normalized to 35%

C. Projects (30%)

Points per project:

• Base Project: 5 points
• Internship Bonus: +5 points (if company is not "personal", "na", or "n/a")
• Technical Relevance: 0-10 points based on:
  • Programming Languages used
  • Frameworks utilized
  • Databases implemented
  • Knowledge areas applied

Score normalized to maximum of 30%

D. Extra-Curricular Activities (15%)

Points allocated based on total number of activities across:

• Leadership Roles
• Awards
• Certifications
• Activities

Scoring scale:

• 0 activities: 0 points
• 1-3 activities: 1 point
• 4+ activities: 2 points

Score normalized to maximum of 15%

3. Data Export

• Structured CSV files for each component
• Raw LLM response in JSON format
• Comprehensive metadata with scores

Setup

1. Install dependencies:

pip install -r requirements.txt

2. Set up environment variables in .env:

DEEPSEEK_API_KEY=your_api_key
DEEPSEEK_URL=your_api_url

3. Run the application:

streamlit run app.py

Output Structure

The system generates the following files under data/parsed_data/{reg_no}/:

1. {reg_no}_metadata.csv: Personal information and overall scores
2. {reg_no}_academic.csv: Semester-wise academic performance
3. {reg_no}_skills.csv: Technical skills breakdown
4. {reg_no}_projects.csv: Project and internship details
5. {reg_no}_extracurricular.csv: Extra-curricular activities

Raw LLM responses are stored in data/raw_responses/{reg_no}/llm_response.json

UI Features

1. Interactive Dashboard

   • Upload interface for PDF resumes
   • Real-time parsing and scoring
   • Visual score breakdown

2. Detailed Visualizations

   • Academic performance trends
   • Score distribution charts
   • Skills and project matrices

3. Data Export

   • Download parsed data in CSV format
   • Access raw LLM responses
   • Export structured metadata

Technical Implementation

1. LLM Integration

   • Uses DeepSeek API for main resume parsing
   • Structured output using Pydantic models
   • Token usage tracking

2. Pattern-Based Extraction

   • Custom extractor for extra-curricular activities
   • Regex-based section identification
   • Smart text cleaning and formatting

3. Scoring System

   • Component-wise score calculation
   • Normalization for each category
   • Weighted aggregation for final score

Prerequisites

• Python 3.8+
• Deepseek API credentials
• PDF files in NITK resume format

Installation

1. Clone the repository:

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

2. Create and activate a virtual environment (recommended):

python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install required packages:

pip install -r requirements.txt

4. Set up environment variables: Create a .env file in the project root with:

DEEPSEEK_API_KEY=your_api_key_here
DEEPSEEK_URL=https://api.deepseek.com

Usage

Option 1: Batch Processing (Command Line)

Use this option when you have a combined PDF containing multiple resumes and want to:

• Split them into individual PDFs
• Extract information from each resume
• Save the data in CSV format

1. Place your combined PDF in the data folder:

mkdir -p data
cp your_combined_resumes.pdf data/resumes_compiled.pdf

2. Run the main script:

python -m resume_parser.main

The script will:

• Split the combined PDF into individual resumes in data/output/pdfs/
• Create parsed data in CSV format in data/output/parsed_data/
• Each student's data will be in a separate folder named by their registration number

Option 2: Interactive Web Interface

Use this option to analyze individual resumes with a visual interface.

1. Run the Streamlit app:

streamlit run app.py

2. Open your browser and go to http://localhost:8501

3. Use the interface to:

• Upload individual resume PDFs
• View parsed information in a structured format
• See academic performance visualizations
• Get skill summaries
• View project details

Setting up Deepseek API

1. Sign up for a Deepseek account at https://deepseek.com

2. Create an API key:

   • Go to your account settings
   • Navigate to API section
   • Generate a new API key

3. Add the API key to your .env file:

DEEPSEEK_API_KEY=your_api_key_here
DEEPSEEK_URL=https://api.deepseek.com

Running Tests

Run the test suite to verify everything is working:

pytest tests/ -v --cov=resume_parser

Architecture