How vishavgupta01/resume-screening-ai-agent works
Project type and purpose
• This is a Python-based resume-screening tool implemented as a Jupyter notebook (Resume_Screening_AI_Agent.ipynb).
• Its purpose is to ingest PDF/DOCX resumes, extract text, compute similarity against a job description, extract years of experience, and produce a ranked list of candidates.
File structure
• Top-level:
• README.md – project description and usage instructions
• Resume_Screening_AI_Agent.ipynb – main processing notebook
• Resumes/ – folder containing sample PDF and DOCX resumes
• No separate frontend, backend, or deployment scripts—this is a standalone data-processing pipeline.
README for additional insights
• Dependencies: python-docx, PyPDF2, nltk, scikit-learn, docx2txt, pandas, numpy
• Workflow described: drop resumes in folder, set job_description_text, run script, view ranked output
Main components and their interactions
a. Resume Ingestion
• Component: File System Reader
• Reads all files in Resumes/ directory (PDF & DOCX)
b. Text Extraction
• Component: PDF Parser (PyPDF2) & DOCX Parser (python-docx or docx2txt)
• Output: raw text per resume
c. Text Preprocessing & Feature Extraction
• Component: NLP Processing (NLTK for tokenization, stop-word removal)
• Component: Vectorizer (scikit-learn TF-IDF or similar)
• Output: numeric vectors for resumes and job description
d. Similarity Scoring
• Component: Similarity Calculator (cosine similarity)
• Inputs: resume vectors, job description vector
• Output: similarity score
e. Experience Extraction
• Component: Experience Parser (regex or NLP rules)
• Output: years of experience per resume
f. Ranking & Output
• Component: Ranking Engine (combines similarity + experience score)
• Output: sorted list printed or saved
Relationships & data flow
→ Two parallel branches:
1. similarity scoring ← job_description_text
2. experience extraction
→ Ranking Engine → Final output
Architectural patterns & design principles
• Pipeline pattern: linear stages, each stage receives data from previous
• Separation of concerns: distinct modules for I/O, parsing, NLP, scoring, ranking
Technologies & libraries
• File I/O: built-in Python file system APIs
• PDF parsing: PyPDF2
• DOCX parsing: python-docx or docx2txt
• NLP: NLTK
• Feature extraction & ML: scikit-learn (TF-IDF, cosine similarity)
• Data handling: pandas, numpy
• Orchestration: Jupyter notebook acts as the driver