An AI-powered resume screening and candidate shortlisting application that automates the recruitment process using machine learning and natural language processing. Built with Streamlit for HR professionals to efficiently manage large volumes of job applications.
- Automated Resume Parsing: Extract structured information from PDF resumes using Groq LLM
- Intelligent Skill Matching: Calculate semantic similarity between candidate skills and job requirements
- Multi-criteria Shortlisting: Filter candidates based on CGPA, experience, and skill compatibility
- Batch Processing: Handle multiple resumes simultaneously with progress tracking
- Interactive Dashboard: User-friendly interface for HR workflow management
- Export Functionality: Download shortlisted candidates as CSV files
- Real-time Processing: Live feedback during resume analysis and shortlisting
- Frontend: Streamlit (Interactive web interface)
- AI/ML:
- Groq LLM (Llama3-8b-8192) for resume parsing
- SentenceTransformers for semantic skill matching
- Scikit-learn for similarity calculations
- Database: SQLite for data persistence
- PDF Processing: PyMuPDF (fitz) for text extraction
- Data Processing: Pandas, NumPy for data manipulation
- Python 3.8 or higher
- Groq API Key (Sign up here)
# Clone the repository
git clone <repository-url>
cd smart-resume-shortlisting
# Install dependencies
pip install streamlit pandas scikit-learn sentence-transformers PyMuPDF groq numpy
# Run the application
streamlit run app.py- Create account at Groq Console
- Generate API key
- Enter API key in the application sidebar
graph TD
A[Start Application] --> B[Initialize SQLite Database]
B --> C[Create Job Post]
C --> D[Upload Resume PDFs]
D --> E[Extract Text with PyMuPDF]
E --> F[Parse with Groq LLM]
F --> G[Store Structured Data]
G --> H[Run Shortlisting Process]
H --> I[Calculate Skill Similarity]
I --> J[Apply Multi-criteria Filter]
J --> K[Update Candidate Status]
K --> L[View Results Dashboard]
L --> M[Export Shortlisted Candidates]
- HR creates job posting with title, description, and requirements
- Define required skills (comma-separated list)
- Set minimum CGPA and experience thresholds
- Store job post in SQLite database
Upload Phase:
- Select target job post for applications
- Upload multiple PDF resumes simultaneously
- Real-time progress tracking during processing
AI Parsing Phase:
PDF Resume → Text Extraction → Groq LLM → Structured JSON
Extracted Data Structure:
{
"full_name": "John Doe",
"email": "john@example.com",
"phone": "+1234567890",
"skills": ["Python", "Machine Learning", "SQL", "Git"],
"cgpa": 8.5,
"experience_years": 3,
"education": [{"degree": "BTech", "university": "XYZ University"}],
"projects": [{"title": "ML Project", "description": "Built recommendation system"}],
"certifications": ["AWS Certified Developer"]
}Multi-criteria Evaluation:
- CGPA Check: candidate_cgpa ≥ minimum_required_cgpa
- Experience Check: candidate_experience ≥ minimum_required_experience
- Skill Similarity: cosine_similarity(candidate_skills, job_skills) ≥ threshold
Similarity Calculation:
- Convert skills to embeddings using SentenceTransformers
- Calculate cosine similarity between candidate and job skill vectors
- Configurable similarity threshold (default: 0.65)
Final Decision Logic:
IF (CGPA ✓ AND Experience ✓ AND Skills ✓)
THEN Status = "Shortlisted"
ELSE Status = "Rejected"
- Summary dashboard with key metrics
- Detailed candidate profiles with scores
- Export shortlisted candidates to CSV
- Individual candidate management (view/delete)
-
Create Job Post
- Job title and detailed description
- Required skills input (comma-separated)
- Minimum CGPA and experience settings
- Form validation and success feedback
-
View Job Posts
- List all created job postings
- Expandable cards with full job details
- Delete functionality for unwanted posts
-
Upload Resumes
- Job post selection dropdown
- Multiple PDF file upload interface
- Real-time processing status updates
- Error handling for failed uploads
-
Shortlist Candidates
- Job selection and requirement display
- Adjustable similarity threshold slider
- Automated evaluation with progress bar
- Success confirmation with metrics
-
View Results
- Summary metrics (Total, Shortlisted, Rejected)
- Detailed candidate cards with all information
- Status indicators (✅ Shortlisted, ❌ Rejected)
- CSV export functionality
CREATE TABLE job_posts (
id INTEGER PRIMARY KEY AUTOINCREMENT,
job_title TEXT NOT NULL,
job_description TEXT NOT NULL,
required_skills TEXT NOT NULL,
min_cgpa REAL NOT NULL,
min_experience INTEGER NOT NULL,
created_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP
)CREATE TABLE candidates (
id INTEGER PRIMARY KEY AUTOINCREMENT,
job_id INTEGER,
candidate_name TEXT NOT NULL,
full_name TEXT,
email TEXT, phone TEXT,
skills TEXT, -- JSON array
cgpa REAL,
experience_years INTEGER,
education TEXT, -- JSON array
projects TEXT, -- JSON array
certifications TEXT, -- JSON array
resume_text TEXT,
similarity_score REAL,
status TEXT,
created_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP,
FOREIGN KEY (job_id) REFERENCES job_posts (id)
)# Groq LLM Settings
model = "llama3-8b-8192"
temperature = 0.1
max_tokens = 4000
# Sentence Transformer
embedding_model = "all-MiniLM-L6-v2"- Skill Similarity Threshold: 0.65 (adjustable via slider)
- Resume Text Limit: 4000 characters for LLM processing
- Batch Processing: Unlimited PDF uploads
1. Create Job Post:
- Title: "Senior Software Engineer"
- Skills: "Python, Django, React, PostgreSQL, AWS"
- Min CGPA: 7.5, Min Experience: 3 years
2. Upload Resumes:
- Upload 25 candidate PDF resumes
- System processes each resume with AI parsing
3. Run Shortlisting:
- Set similarity threshold to 0.7
- System evaluates all candidates automatically
- Results: 8 shortlisted, 17 rejected
4. Export Results:
- Download shortlisted_candidates_20241201_143022.csv
- Share with hiring team for interviews
- PDF text extraction failures with user feedback
- Groq API connection issues and retry mechanisms
- JSON parsing errors with fallback processing
- Database connection problem resolution
- Invalid file format detection and warnings
Installation Issues:
# If PyMuPDF installation fails
pip install --upgrade pip
pip install PyMuPDF
# If sentence-transformers has issues
pip install torch torchvision
pip install sentence-transformersRuntime Issues:
- Groq API Errors: Verify API key and internet connection
- Database Locked: Restart application if SQLite locks occur
- Memory Issues: Process resumes in smaller batches for large datasets