This project is an end-to-end system that parses resumes in PDF format, extracts structured information, and matches candidate profiles against job descriptions using semantic similarity. The final output is a detailed Excel report ranking resumes based on how well they fit the job requirements.
Recruiters often receive thousands of resumes per role. Manually screening them is time-consuming and error-prone. This system automates:
- Resume parsing (education, experience, projects, etc.)
- College tier classification (Tier 1/2/3)
- Project extraction
- Semantic job-resume matching
- Outputting an Excel report with job-fit probability
- Python
pdfplumberorPyMuPDFfor PDF parsingfuzzywuzzyfor fuzzy string matchingpandasfor data processingsentence-transformers(BERT) for semantic similarityopenpyxlfor Excel export
- Parses resumes from a folder of PDFs
- Extracts key fields: Institute, Year, CGPA, Branch, Experience, Projects
- Maps colleges to tiers using fuzzy matching
- Uses BERT to compute job fit score
- Exports all results to Excel
Chosen for accurate and lightweight extraction of text from PDFs. It works well for standard resume layouts.
Used to robustly match institute names from resumes to those in the college tier dataset. This handles abbreviations, typos, and formatting inconsistencies.
BERT was chosen because it captures the meaning of text, not just keywords. This allows for accurate matching between resumes and job descriptions, even when phrased differently.
Machine learning models like Random Forest require large labeled datasets. Since we don't have resume/job outcome labels, a pre-trained model like BERT is more suitable and ready-to-use.
- Extract text from PDF resumes
- Parse fields using regex and string rules
- Map institutes to tiers using fuzzy matching
- Load job description from CSV
- Use BERT to compute cosine similarity between resume and JD
- Store similarity as Job_Fit_Probability (%)
- Export to Excel with all extracted fields
Developed by Aaisha Rani
Feel free to connect on: