SELECTRA Documentation
Project Overview
SELECTRA is an AI-powered resume screening application designed to streamline the hiring process for HR teams. It uses advanced natural language processing to analyze resumes against job requirements, ranking candidates based on their suitability.
Why We Built SELECTRA
The traditional resume screening process is time-consuming and often subjective. SELECTRA addresses these challenges by:
- Automating the initial screening of hundreds of resumes
- Providing objective, consistent evaluation based on job requirements
- Identifying top candidates with quantifiable match scores
- Generating targeted interview questions based on candidate profiles
- Saving HR teams countless hours in the hiring process
Development Team
SELECTRA is developed by the students of Software Engineering Department of SMIU.

Wasif Mehmood Ali
Team Lead / Backend / AI

Shaharayar Khan
UI/UX Lead

Nasarullah
Frontend Developer

Tajammul Abbasi
UI & Frontend Contributor

Zain Zaib
Research Lead
Technical Pipeline
The SELECTRA application follows a structured pipeline to process resumes and generate insights:
Resume Upload
Users upload resumes in various formats (PDF, DOCX, TXT) through a drag-and-drop interface or file browser. The application supports batch processing of multiple resumes simultaneously.
Text Extraction
The application extracts raw text from uploaded documents using format-specific libraries:
- PDF files: Processed using PDF.js, a Mozilla-developed library that parses PDF documents in the browser without server-side dependencies.
- DOCX files: Processed using Mammoth.js, which converts DOCX documents to HTML and then extracts the text content.
- TXT files: Read directly using the FileReader API.
AI Analysis with Gemini
The extracted text is sent to Google's Gemini API with a carefully crafted prompt that includes:
- The job requirements entered by the HR team
- Instructions for analyzing the resume content
- A specific response format for consistent parsing
The API returns structured data including candidate name (extracted from resume), match score (0-100), relevant experience, key skills, and education.
Results Processing
The application processes the API response to:
- Sort candidates by match score (descending)
- Calculate aggregate metrics (average score, top candidates, etc.)
- Generate visual indicators (score bars, color-coded ratings)
- Prepare data for display in the results table
Interview Questions Generation
For the top candidates, SELECTRA generates targeted interview questions by:
- Sending the job requirements and candidate profiles to Gemini
- Requesting specific, relevant questions categorized by type (technical, behavioral, etc.)
- Formatting the responses for clear display in the UI
Export Functionality
HR teams can export results in multiple formats:
- PDF Export: Uses jsPDF to generate professional reports with candidate rankings and interview questions
- Excel Export: Uses SheetJS to create spreadsheet-ready data for further analysis
Evaluation Methodology
SELECTRA evaluates candidates based on several factors extracted from their resumes:
Match Score Calculation
The AI evaluates each resume against the job requirements and assigns a score (0-100) based on:
- Relevant Experience: Years and quality of experience matching the job requirements
- Skills Match: Presence and proficiency of required skills
- Education: Relevance and level of education
- Keyword Analysis: Presence of important keywords from the job description
Score Categories
- 90-100% Excellent match - Strong in all required areas
- 80-89% Very good match - Meets most requirements
- 70-79% Good match - Meets many requirements
- 60-69% Fair match - Meets some requirements
- Below 60% Weak match - Lacks key requirements
Interview Questions Generation
Questions are tailored based on:
- Gaps between candidate qualifications and job requirements
- Areas where more information is needed
- Specific skills or experiences mentioned in the resume
- Common interview questions for the role type
Technical Stack
SELECTRA is built with modern web technologies: