Transparent • Fair • Explainable Recruitment powered by AI
AI in hiring should not just be accurate — it should be fair, transparent, and accountable.
FairHire is an Explainable AI (XAI)-driven recruitment system that transforms hiring into a data-driven, bias-aware, and interpretable process.
Modern AI hiring systems often:
- ❌ Operate as black boxes
- ❌ Hide decision logic
- ❌ Introduce bias (gender, education, etc.)
- ❌ Reduce recruiter trust
➡️ Result: Unfair and unreliable hiring decisions
FairHire is an end-to-end intelligent hiring system that:
- 🤖 Predicts candidate hiring probability
- 🧠 Explains decisions using SHAP
- ⚖️ Detects bias using fairness metrics
- 📄 Extracts resume data using NLP
- 📊 Visualizes insights via interactive dashboard
👉 Not replacing recruiters — empowering them
- Random Forest Classifier
- Predicts probability of selection
- Shows why a candidate was selected
- Feature-level contribution
- Transparent decision-making
- Uses Demographic Parity
- Detects imbalance across groups
- Flags bias automatically
-
PDF parsing via
pdfplumber -
NER using BERT
-
Extracts:
- Skills
- Experience
- Certifications
- Real-time filtering
- Candidate comparison
- Hiring probability insights
- Bias monitoring
flowchart TD
A[Resume Input] --> B[Data Processing]
B --> C[Feature Extraction]
C --> D[Random Forest Model]
D --> E[SHAP Explainability]
D --> F[Fairness Engine]
E --> G[Dashboard]
F --> G
G --> H[Human Decision Support]
| Metric | Purpose |
|---|---|
| Accuracy | Overall correctness |
| Precision | Quality of shortlist |
| Recall | Identifying good candidates |
| ROC-AUC | Model separability |
✔ Balanced performance across hiring decisions ✔ Reliable classification with explainability
- 📌 Interview score & experience drive hiring decisions
- 📌 SHAP reveals clear decision reasoning
- 📌 Bias detection ensures fairness across groups
- 📌 Dashboard enables data-driven HR decisions
Language:
- Python
ML & Data:
- Scikit-learn
- Pandas, NumPy
Explainable AI:
- SHAP
NLP:
- BERT (dslim/bert-base-NER)
- pdfplumber
Visualization:
- Plotly, Matplotlib
Deployment:
- Streamlit
- Pickle
git clone https://github.com/aragrishah/FairHire_Project.git
cd FairHire_Project
pip install -r requirements.txt
streamlit run app.py- Synthetic dataset
- Limited real-world hiring signals
- No enterprise HR integration
- Deep Learning-based scoring
- Real-time hiring pipelines
- Cloud deployment (AWS/GCP)
- Bias mitigation algorithms
- Interview AI integration
- 🌍 Solves real-world ethical AI problem
- 🧠 Combines ML + NLP + XAI + Fairness
- ⚖️ Focus on Responsible AI
- 📊 Strong product-level dashboard
- 💼 Highly relevant for AI/Data roles
- Riya Shah
- Jhanvi Vakharia
FairHire proves that AI can be both powerful and responsible — not just predicting outcomes, but explaining them.