Resume Parser & Job Matching

A project that extracts and summarizes PDF resumes using both code-based logic (regex, dictionaries) and LLM refinement (Groq LLM). It then compares the finalized resume data against a user-provided job description using lemma-based and semantic matching approaches, yielding detailed matching scores.

Live Demo

Project Deployment Link: Resume Parser App

Video Demonstration

Check out the demo video here: Demo Video

Features

1. PDF Resume Upload:
   Upload any PDF resume file.

2. Code-Based Parsing:

   • Regex for phone and email.
   • List-based skill detection from known keywords.

3. LLM Refinement:

   • A Groq LLM verifies partial parsed data, removing incorrect items and producing a final summary in subpoints (Education, Experience, Skills, etc.).
   • Fallback text appears if the LLM or API key is unavailable.

4. Job Description Matching:

   • Lemma-Based (Jaccard / lexical overlap).
   • Semantic (using sentence-transformers to measure embedding similarity).
   • Combined final score.

5. Detailed Outputs:

   • Raw vs. Cleaned Resume Text
   • Matched & Unmatched Tokens / Sentences
   • Final Summaries / Bullet Points

How It Works

1. User Interaction:

   • Paste a job description.
   • Upload a PDF resume.

2. Partial Parsing:

   • Regex extracts phone/email.
   • Keyword detection finds known skills (e.g., "Python", "Java").

3. LLM Finalization:

   • The partial parse plus the raw text is fed into Groq LLM.
   • The LLM verifies or removes incorrect fields, then produces subheading-based summaries.

4. Matching with JD:

   1. Lemma-Based: Jaccard overlap between lemmatized tokens from the JD and resume.
   2. Semantic: Overall text similarity plus line-by-line JD comparisons.
   3. Combined: Weighted average, default 50/50.

Local Deployment

1. Clone or Download the repository.
2. Install the dependencies (pinned in requirements.txt):

   pip install -r requirements.txt

3. Run the Streamlit app:

   streamlit run app.py

4. Open the URL provided by Streamlit.
5. Interact with the UI to upload a PDF resume and paste a job description.

Project Flow

1. app.py:

   • Presents the Streamlit UI.
   • Handles user input (resume + JD).
   • Shows partial parse, LLM finalization, and matching outputs.

2. requirements.txt:

   • Lists pinned versions for stable deployment.

3. Code & LLM synergy:

   • The code-based approach ensures partial data extraction without relying solely on the LLM.
   • The LLM refines that data, producing a final bullet-point summary.

Demo Links

• Live App: Resume Parser App
• Demo Video: Watch on Google Drive

Known Limitations

• If the Groq LLM is unavailable or the key is invalid, you'll see a fallback message.
• The code-based parse is minimal (phone/email regex, skill dictionary). Extend these methods for deeper extraction.
• Torch-based dependencies can occasionally cause environment conflicts. See requirements.txt for pinned versions.

Contributing

1. Fork this repository.
2. Create a new branch for your features/fixes.
3. Open a Pull Request with a clear explanation.

License

This project is open-source under the MIT License. Feel free to use and adapt it.

Acknowledgements

• Streamlit for interactive UI.
• PyPDF2 for PDF text extraction.
• Groq for LLM inference.
• NLTK & sentence-transformers for textual processing & semantic matching.