How hazzanzz/ai-powered-resume-ranker works
Project type and purpose
• This is a lightweight full-stack application built with Streamlit (Python) that ranks and improves resumes.
• It ingests two inputs (resume PDF, job description TXT), applies NLP (SBERT, YAKE), and calls an external AI service (OpenAI ChatGPT) to produce scores, keyword comparisons, and improvement suggestions.
File structure
• app.py
• The Streamlit entry point (frontend + controller).
• resume_matcher.py
• Encapsulates semantic similarity logic (SBERT).
• keyword_extractor.py
• Encapsulates YAKE keyword extraction.
• chatgpt_improver.py
• Wraps OpenAI ChatGPT API calls for improvement suggestions.
• job_description.txt, sample_data/randomized-resume.pdf
• Sample inputs.
• test_keywords.py, test_similarity.py
• Unit tests for core NLP modules.
• requirements.txt
• Lists Python dependencies.
• secret.gitignore
• Hides API keys.
README for additional insights
• Tech stack: Streamlit frontend, SBERT, YAKE, PyPDF2, OpenAI API.
• Features: PDF parsing → semantic similarity → keyword extraction → ChatGPT suggestions.
Main components & their interactions
a. Components
1. User Interface (Streamlit app.py)
2. PDF/Text Ingestion (PyPDF2 in app.py)
3. Keyword Extractor (keyword_extractor.py with YAKE)
4. Semantic Similarity Processor (resume_matcher.py with SBERT)
5. AI Improver (chatgpt_improver.py calling OpenAI API)
6. Data Stores (in-memory or local file I/O for inputs; no persistent DB)
7. External Service (OpenAI ChatGPT API)
b. Interactions
• User uploads resume & job description → Streamlit UI
• app.py invokes PDF/Text Ingestion → raw text
• Text passes in parallel to Keyword Extractor and Semantic Similarity Processor
• Results aggregated (match % + top keywords)
• If user requests improvements, app.py calls AI Improver → OpenAI API → suggestions
• Final output rendered back to Streamlit UI
c. Architectural patterns / design principles
• Modular separation: each NLP task is encapsulated in its own module
• “Controller” (app.py) orchestrates data flow
• Stateless processing: each request runs pipeline from scratch
• External dependency isolation: secrets in .gitignore
d. Key technologies
• Streamlit (frontend & minimal server)
• PyPDF2 (PDF parsing)
• SBERT via sentence-transformers (semantic similarity)
• YAKE (keyword extraction)
• OpenAI API (ChatGPT enhancements)
• pytest (unit tests)
Detailed elements
• Component names (e.g., app.py: Streamlit Controller)
• Data stores/inputs (resume PDF, JD TXT)
• OpenAI ChatGPT with API key (secret management)