AI-Powered Resume Screening System

The AI Powered Resume Screening System is a full-stack HR automation project built with Python, Flask, spaCy, and scikit-learn that solves a genuine business problem: manually reviewing hundreds of resumes for a single job opening. Recruiters upload resumes in PDF or DOCX format, define a job description with required skills, minimum experience, and education level, and the system parses every file, extracts structured candidate information, and ranks applicants by a weighted Job-Match Score percentage. Unlike a basic keyword tool, it uses TF-IDF and cosine similarity for semantic matching, with an optional BERT engine for deeper contextual understanding. The result is an enterprise-grade HR-Tech application combining NLP, machine learning, document parsing, REST APIs, and data visualisation.

Key Features

• The resume parser extracts clean text from messy multi-column PDF and DOCX files using PyMuPDF and python-docx.
• The NLP extractor pulls candidate name, email, phone, skills, education, and experience using spaCy Named Entity Recognition combined with regex patterns.
• The job description module lets you define required skills, minimum years of experience, and education level, with skills auto-detected from the JD text.
• The weighted match engine scores each candidate across skills, experience, education, and semantic similarity using TF-IDF or optional BERT.
• An interactive analytics dashboard built with Chart.js shows ranking bar graphs, score distribution doughnut charts, and per-candidate skill-gap analysis.
• One-click shortlist and reject actions update candidate status over AJAX with live filtering.
• Ranked or shortlisted candidates can be exported as CSV or Excel files using openpyxl.
• The database is pluggable: SQLite works out of the box and MySQL is enabled with a single environment variable change.
• Score weights are fully configurable in the config file, and pytest test cases are included for the core matching engine.

Technology Stack

System Requirements

• Python 3.x with pip installed.
• Flask and SQLAlchemy for the web application and ORM layer.
• spaCy with the en_core_web_sm English model (the app falls back to regex extraction without it).
• scikit-learn for TF-IDF vectorization and cosine similarity.
• PyMuPDF and python-docx for reading PDF and Word resumes.
• openpyxl for Excel export and pytest for running the test suite.
• SQLite by default, or MySQL with pymysql for production setups.
• Optional sentence-transformers package if you want to enable the BERT match engine.

How It Works

A recruiter creates a job description, and the system auto-detects the required skills from the text along with the minimum experience and education level. Uploaded resumes are converted to plain text by PyMuPDF or python-docx, then passed to the spaCy NER pipeline which extracts the candidate's name, contact details, skills, education, and years of experience. The match engine blends four signals into one score: skills overlap at 50 percent, experience at 20 percent, education at 15 percent, and TF-IDF or BERT semantic similarity at 15 percent. Every candidate is ranked by this percentage on the dashboard, where the recruiter can review skill gaps, shortlist or reject with one click, and export the final list.

Perfect For

This is an advanced-level project suited to MCA, BCA, B.Tech CS/IT, M.Tech, and MBA IT or HR students who want a final year submission that stands out. It covers artificial intelligence, natural language processing, machine learning, and full-stack web development in a single codebase, giving you plenty of material for viva questions on NER, TF-IDF, cosine similarity, and transformer models. It also doubles as a strong portfolio and interview project for AI, Data Science, and HR-Tech roles.

Future Enhancements

The project can be extended with recruiter and admin login with role-based access, automated email notifications to shortlisted candidates, resume ranking across multiple jobs at once, an interview scheduling module, bulk ZIP upload of resumes, cloud deployment on Render or Railway, and a fine-tuned transformer model for domain-specific skill extraction.

Watch Project Demo

Watch the full project walkthrough and setup tutorial on our YouTube channel.

▶ Watch on YouTube – Decode It

More Details & Full Documentation

For complete project details, source code, and setup guide, visit the official project page.

🔗 View Full Project Details – UpdateGadh