Project Overview
The AI Resume Screening & Ranking System is a machine learning–powered Applicant Tracking System (ATS) designed to assist recruiters in automatically analyzing and ranking job applicants based on their resumes. The system processes uploaded resumes, extracts relevant information, and predicts the most suitable job category using trained classification models. This helps automate the early stages of recruitment and significantly reduces manual resume screening efforts.
In addition to classification, the system implements an intelligent resume ranking mechanism that compares multiple resumes against a given job description. By applying Natural Language Processing techniques such as TF-IDF vectorization and cosine similarity, the system measures how closely each candidate’s profile matches the job requirements. This allows recruiters to quickly identify the most relevant candidates for a particular role.
To provide a user-friendly experience, the system includes a FastAPI backend for model inference and a recruiter dashboard interface that enables uploading resumes, viewing predictions, and ranking candidates in real time. The platform integrates multiple scoring factors such as similarity score, predicted job category confidence, and experience extraction to generate a hybrid ATS ranking score, making candidate evaluation more accurate and efficient.
Key Features
- Resume classification using machine learning models
- Automatic resume text extraction from uploaded PDF files
- Multi-resume candidate ranking against job descriptions
- Hybrid ATS scoring combining similarity, classification confidence, and experience
- Natural Language Processing using TF-IDF vectorization
- Cosine similarity for job description matching
- Experience extraction from resume text
- FastAPI REST API for model inference
Technical Approach
The system begins with data preprocessing and feature engineering using a dataset containing more than 2,200 labeled resumes across various job categories. Resume text is cleaned and transformed using Natural Language Processing techniques. TF-IDF vectorization is applied to convert textual resume content into numerical feature vectors that can be used by machine learning models.
For classification, the project uses machine learning algorithms such as Random Forest and XGBoost to predict the job category of a resume. The models are trained and evaluated on the processed dataset to learn patterns associated with different professional roles. Once trained, the best performing model is deployed as part of the inference pipeline to classify new resumes uploaded by users.
To enhance recruitment decision-making, the system implements a hybrid candidate ranking mechanism. Multiple resumes are compared with a job description using TF-IDF cosine similarity to measure textual relevance. The similarity score is then combined with classification confidence and extracted experience to produce a final ATS score. This weighted scoring approach allows the system to rank candidates more intelligently and provide recruiters with an ordered list of the most suitable applicants.