FairLens is a cloud-based AI platform designed to promote diversity, equity, and inclusion (DEI) in hiring. It uses advanced machine learning models and LLMs to analyze resumes, detect biases, and provide fair candidate shortlisting — all while respecting privacy through PII anonymization.
- 🤖 AI-Powered Resume Screening using LLMs
- 📊 Real-Time Diversity Dashboard (Dash-based UI)
- 🔒 PII Anonymization for candidate data security
- 🎯 Custom Diversity Goal Setup for companies
- ☁️ Secure Cloud Infrastructure with AWS Cognito, S3, and DynamoDB
- 💸 Cost-Optimized Resume Parsing using GPT-4o Mini (~$0.000075 per resume)
- Languages: Python, HTML, CSS, JavaScript
- Frameworks/Libraries: Flask, Dash, Pandas, Scikit-learn, XGBoost, SVR, LangChain
- Cloud: AWS Cognito, S3, DynamoDB, Boto3
- AI & ML: OpenAI GPT-4o Mini, Random Forest, XGBoost, SVR
- DevOps & Deployment: Modular MVC structure, secure AWS hosting
FairLens is built on a robust, scalable, and secure AWS infrastructure that supports its data processing, storage, and authentication needs.
| Service | Purpose |
|---|---|
| Amazon Cognito | User authentication and identity management |
| Amazon S3 | Scalable storage for resumes and static assets |
| Amazon DynamoDB | NoSQL database for storing structured company and job data |
| AWS Boto3 SDK | Python SDK for seamless interaction with AWS services |
This architecture ensures seamless integration, low-latency performance, and secure handling of sensitive candidate data.
fairlens/
├── app.py
├── dashboard/ # Dash UI
├── anonymizer/ # PII Anonymization logic
├── resume_parser/ # Resume handling and ML logic
├── auth/ # AWS Cognito integration
├── utils/ # Helper modules
├── static/ # CSS/JS files
├── templates/ # HTML templates
├── requirements.txt
└── README.md
Update the following placeholders in your .env or config file:
OPENAI_API_KEY=“<your-openai-api-key>”
COGNITO_REGION="us-east-1"
COGNITO_USER_POOL_ID="<your-user-pool-id>"
COGNITO_CLIENT_ID="<your-client-id>"
COGNITO_CLIENT_SECRET="<your-client-secret>"git clone https://github.com/<your-username>/fairlens.git
cd fairlens
pip install -r requirements.txt
python app.py- Company Onboarding: Register your organization and set diversity goals.
- Resume Submission: Upload resumes and job descriptions.
- Resume Parsing & Anonymization: All resumes are anonymized before analysis.
- AI Shortlisting: Candidates are matched using LLMs + diversity-weighted scoring.
- Diversity Dashboard: Monitor team composition and hiring pipeline health.
- Live DEI statistics by role, department, and funnel stage
- Predictive analytics for attrition, gender ratio, and performance
- AI-generated insights for recruitment policy improvements
| Model | Algorithm | Purpose |
|---|---|---|
| Attrition Prediction | XGBoost | Predicts risk of employee attrition |
| Gender/Diversity Forecast | Random Forest | Tracks inclusion metrics over time |
| Team Performance | SVR | Evaluates team performance vs. size |
| Resume Screening | LLM + LangChain | Identifies top candidates & removes bias |
- ⚖️ Reduced unconscious bias in screening
- 👥 Increased underrepresented hires (by skill match)
- 📉 Cost reduced to ~$0.000075/resume using GPT-4o Mini
- 🔐 PII anonymization ensures candidate privacy
FairLens uses dynamic weighting logic based on:
- Province-level diversity data (e.g., Statistics Canada)
- User-defined hiring objectives
- Bias detection heuristics & LLM transparency
For example:
If the average representation of women in Manitoba is 48% compared to the national 55%, the diversity weight for women will be: 55/48 = 1.15.
- Integrate with LangChain agents for job-specific reasoning
- Add feedback loops to fine-tune AI shortlisting
- Expand to multilingual resume processing
- Hetkumar Dineshbhai Patel
- Bhagyesh Jayeshbhai Patel
- Parth Jigneshkumar Patel
Supervisor: Prof. Kobra Khanmohammadi
For questions or demo access:
FairLens — Because every qualified candidate deserves a fair shot. 🌍
Before running the project, install the required Python packages:
pip install -r requirements.txtFairLens utilizes a suite of AWS cloud services to ensure scalability, security, and performance:
| AWS Service | Purpose |
|---|---|
| Amazon Cognito | User authentication and profile management |
| Amazon DynamoDB | Structured data storage (job roles, company info, etc.) |
| Amazon S3 | Storage for resume documents and other unstructured data |
| AWS Boto3 SDK | Python interface for communication with AWS services |
+------------------+
| Company Registers |
+--------+---------+
|
v
+------------------------+
| Set Diversity Goals |
+--------+---------------+
|
v
+------------------------+
| Create Job Description |
+--------+---------------+
|
v
+------------------------+
| Upload Resumes |
+--------+---------------+
|
v
+------------------------------+
| Resume Parsing & Anonymizer |
+--------+---------------------+
|
v
+------------------------------+
| AI Screening (LLMs + ML) |
+--------+---------------------+
|
v
+------------------------------+
| Candidate Shortlist |
+--------+---------------------+
|
v
+------------------------------+
| Diversity Dashboard & Report |
+------------------------------+
FairLens empowers organizations to make informed hiring decisions with transparency and accountability, combining fairness and performance at every step.