Getting Started with VeraScope
Complete your first bias audit in under 24 hours
Quick Start Timeline
5m
Account Setup
Sign up and get credentials
30m
Model Preparation
Export model and configure audit
24h
Audit Completion
Receive comprehensive bias report
1
Request Access
Create an account to receive container access and credentials for running the audit in your environment.
- Create your account
- Verify your email address
- Receive your API keys and container access
2
Download Audit Container
Pull our secure Docker container that runs in your infrastructure.
- Authenticate with provided credentials
- Pull the latest audit container image
- Verify the image with Cosign signature
- Review included documentation and examples
# Authenticate with VeraScope registry
docker login registry.verascope.ai
# Pull the audit container
docker pull registry.verascope.ai/audit:latest
# Verify signature (optional but recommended)
cosign verify registry.verascope.ai/audit:latest3
Prepare Your Model
Export your model in a supported format with test data.
- Export to ONNX (recommended) or use native format
- Prepare demographic test data (or use our synthetic cohorts)
- Define fairness constraints and protected attributes
- Configure audit parameters in config.yaml
# Example: Export sklearn model to ONNX
import onnx
from skl2onnx import convert_sklearn
onnx_model = convert_sklearn(
your_model,
initial_types=[('input', FloatTensorType([None, n_features]))]
)
onnx.save_model(onnx_model, 'model.onnx')
# Create config.yaml
cat > config.yaml << EOF
model_path: ./model.onnx
protected_attributes: ['gender', 'race', 'age_group']
fairness_constraints:
demographic_parity_difference: 0.1
equal_opportunity_difference: 0.1
EOF4
Run Bias Audit
Execute the audit container with your model and configuration.
- Mount your model and data directories
- Run the container with appropriate resources
- Monitor progress through real-time logs
- Audit completes in under 24 hours
# Run the audit
docker run -v $(pwd):/workspace \
registry.verascope.ai/audit:latest \
--config /workspace/config.yaml \
--output /workspace/results/
# Monitor progress
docker logs -f <container_id>
# Results will include:
# - results/audit_report.pdf (FDA-style report)
# - results/metrics.json (machine-readable)
# - results/recommendations.md (action items)5
Review Results & Next Steps
Analyze your bias audit results and plan remediation.
- Review PDF report for executive summary
- Examine detailed metrics in JSON output
- Identify coverage gaps for synthetic data
- Plan next steps based on recommendations