Project Overview
This system was developed over 16 weeks as part of the FIU Senior Design Capstone program. It addresses a real-world problem in the hiring process: the time and inconsistency involved in manually screening large volumes of resumes against job requirements.
Traditional resume screening relies on keyword matching, which fails to capture actual candidate fit. This system uses a large language model to read and understand both the resume and job description, evaluating experience, skills, and qualifications in the same way a trained recruiter would.
System Pipeline
PDF UploadUser interface
→
Text ExtractionPyPDF2
→
AI EvaluationAnthropic API
→
JSON ParsingStructured output
→
Ranked ResultsSorted by score
- PDF Ingestion: Resumes are uploaded and text is extracted page by page using PyPDF2.
- AI Evaluation: Extracted text is sent to Claude alongside the job description. A structured prompt instructs the model to evaluate actual fit, not keyword overlap.
- Structured Output: The model returns a JSON object containing a score, match label, matched and missing skills, strengths, gaps, recommendations, and a hiring recommendation.
- Ranking: Multiple candidates are scored and sorted automatically from highest to lowest fit.
Technology Stack
Python 3
Core backend language
Flask
Web application framework
Anthropic API
AI language model
PyPDF2
PDF text extraction
HTML / CSS / JS
Frontend interface
python-dotenv
Secure config management
Key Features
- AI-Powered Fit Scoring (0-100): Each resume is scored based on genuine role fit, not keyword frequency.
- Batch Resume Screening: Upload up to 10 resumes at once and receive ranked results side by side.
- Structured Analysis: Every candidate receives a breakdown of matched skills, missing skills, strengths, gaps, and actionable recommendations.
- Hiring Recommendation: Strong Recommend, Recommend, Borderline, or Do Not Recommend with a written justification.
- Sample Job Descriptions: Built-in presets for common roles to enable rapid demonstrations.
- CSV Export: Download ranked results as a spreadsheet.
- Session History: Track previous screenings within a session.
- Drag and Drop Upload: Supports multiple PDF files simultaneously.
Development Team
Developed by the Capstone II project team at the Knight Foundation School of Computing and Information Sciences, Florida International University, Spring 2026.