How bot041/resumescreening-agent-ml-model works
Resume Screening Agent ranks resumes against a job description through its CLI or Python API: it parses PDF, DOCX, or TXT files, extracts structured match features, computes lexical and semantic similarity, and predicts and orders relevance scores. The separate training path generates synthetic examples, builds features, fits an XGBoost model, and saves artifacts used at inference. The README documents both entry modes and CSV output. Training internals are only partly sampled, so the graph limits that branch to documented behavior.
Main entry points: rank_candidates.py Ranker / CLI · train_model.py Training pipeline · parse_resume.py Resume parser · extract_features.py Feature extraction
Components 14
Ranking workflow
- Ranker / CLI
src/rank_candidates.py - Job description
- Resume files
- Ranked results
- CSV output
Resume analysis
- Resume parser
src/parse_resume.py - Feature extraction
src/extract_features.py - Text and taxonomy
src/utils.py - Similarity scorer
src/similarity.py - Sentence-BERT model
Model training
- Training pipeline
src/train_model.py - Synthetic examples
Model artifacts
- Model and scaler
models
Other components
- Recruiter / caller
Connections 19
- Recruiter / caller → Ranker / CLI: invokes
- Recruiter / caller → Job description: provides
- Recruiter / caller → Resume files: provides
- Ranker / CLI → Resume parser: parses folder
- Resume files → Resume parser: reads files
- Resume parser → Ranker / CLI: returns text
- Ranker / CLI → Similarity scorer: scores pairs
- Similarity scorer → Sentence-BERT model: encodes text
- Ranker / CLI → Feature extraction: extracts features
- Feature extraction → Text and taxonomy: extracts metadata
- Job description → Ranker / CLI: supplies text
- Ranker / CLI → Model and scaler: loads model and scaler
- Ranker / CLI → Ranked results: produces ranking
- Ranker / CLI → CSV output: writes CSV
- Recruiter / caller → Training pipeline: starts training
- Training pipeline → Synthetic examples: generates examples
- Training pipeline → Similarity scorer: computes features
- Training pipeline → Feature extraction: builds features
- Training pipeline → Model and scaler: saves artifacts