For AI/ML Teams

Ship Compliant AI Without Slowing Down

Integrate compliance into your ML workflows, not around them. Governum provides the tooling AI engineers need to meet EU AI Act requirements while maintaining development velocity.

See Technical Demo View Integrations
governum.yaml Valid 
ai_system:
  name: "fraud-detection-v2"
  risk_category: "high"
  annex_iii_ref: "5(b)"

documentation:
  technical_spec: "./docs/tech-spec.md"
  risk_assessment: "./docs/risk.md"

monitoring:
  drift_threshold: 0.05
  performance_metrics:
    - accuracy
    - fairness_score
  alert_channels:
    - slack: "#ml-alerts"

Built for Developer Experience

Compliance as code, not compliance as bureaucracy

CLI First

Register systems, run compliance checks, and generate docs from your terminal.

Git Integration

Version control for compliance docs. PR-based review workflows for changes.

API Native

Full REST & GraphQL APIs. Integrate compliance checks into any workflow.

SDK Support

Python, JavaScript, and Go SDKs for programmatic compliance management.

Fits Your ML Stack

Native integrations with the tools you already use

🔬
MLflow
🧪
Weights & Biases
🚀
Kubeflow
📊
SageMaker
☁️
Vertex AI
🔷
Azure ML

CI/CD Pipeline Integration

Add compliance gates to your ML pipelines. Automated checks ensure models can't reach production without proper documentation and risk assessment.

  • GitHub Actions workflow
  • GitLab CI templates
  • Jenkins plugins
  • Azure DevOps extensions
.github/workflows/ml-deploy.yml 
- name: Governum Compliance Check
  uses: governum/compliance-action@v2
  with:
    api-key: ${{ secrets.GOVERNUM_KEY }}
    system-id: fraud-detection-v2

- name: Generate Documentation
  run: governum docs generate

- name: Upload Artifacts
  run: governum artifacts push ./model

Model Registry Sync

Automatic synchronization with your model registry. Every model version is tracked with its corresponding compliance documentation.

  • Auto-discovery from registry
  • Version-linked documentation
  • Deployment tracking
  • Lineage preservation
Model Versions
Version Stage Compliance
v2.3.1 Production Complete
v2.4.0 Staging Review
v2.5.0-rc1 Dev Incomplete

Documentation That Writes Itself

Stop context-switching between code and compliance docs. Governum auto-generates technical documentation from your codebase, model metadata, and training artifacts.

Code Analysis

Extract model architecture, dependencies, and data flows from source code.

Data Lineage

Automatic tracking of training data sources, transformations, and quality metrics.

Performance Capture

Import metrics from your experiment tracking tools automatically.

Auto-Generated Documentation
Model Architecture
Extracted from model.py
Auto
Training Data Summary
Extracted from data pipeline
Auto
Performance Metrics
Synced from MLflow
Auto
Intended Use Description
Manual input required
Manual

ML Observability + Compliance

Monitor model performance and compliance status together

Drift Detection

Monitor data drift and model drift in real-time. Automatic alerts when drift exceeds compliance thresholds.

drift_threshold: 0.05
Fairness Metrics

Track fairness metrics across protected attributes. Built-in support for demographic parity, equalized odds, and more.

fairness_score: 0.92
Decision Logging

Automatic logging of model predictions with full context. Queryable audit trail for regulatory requirements.

retention: 10years
Explainability

Integration with SHAP, LIME, and other explainability tools. Generate explanations that satisfy transparency requirements.

explainer: shap
Alert Integration

Compliance alerts to Slack, PagerDuty, or any webhook. Configure severity levels and escalation paths.

slack: #ml-alerts
Human-in-the-Loop

Configure human oversight requirements. Automatic routing for review when confidence is low.

review_threshold: 0.7

Python SDK

Integrate compliance checks directly into your training scripts. A few lines of code to ensure every model is compliant.

View Full Documentation
train.py Python 
from governum import GovernumClient, AISystem

# Initialize client
client = GovernumClient(api_key=os.environ["GOVERNUM_KEY"])

# Register or update AI system
system = client.systems.upsert(
    id="fraud-detection-v2",
    name="Fraud Detection Model",
    risk_category="high",
    purpose="Detect fraudulent transactions",
    owner="ml-team@company.com"
)

# Log training run
with system.training_run() as run:
    run.log_dataset(train_data, name="training_set")
    run.log_params(model.get_params())

    model.fit(X_train, y_train)

    run.log_metrics({
        "accuracy": accuracy_score(y_test, y_pred),
        "fairness": fairness_score(y_test, y_pred, sensitive)
    })

# Check compliance before deployment
result = system.check_compliance()
if not result.passed:
    raise ComplianceError(result.issues)

Ready to Ship Compliant AI?

Join ML teams at leading companies building responsible AI with Governum.

Schedule Technical Demo View on GitHub