Model Governance and Explainable AI as tools for legal compliance and risk management

How do you prove a 'black box' model is fair and compliant? Learn how model governance and explainable AI provide the tools for risk management and auditing.

#1about 3 minutes

The challenge of operationalizing production machine learning systems

An AI-powered recruiting tool example illustrates the risks and complexities of deploying machine learning models beyond the notebook.

#2about 2 minutes

Four pillars for deploying successful machine learning systems

Successful long-term ML deployment requires a combination of MLOps, model governance, data governance, and explainable AI.

#3about 4 minutes

Understanding model governance and emerging legal frameworks

Model governance addresses legal compliance, like the EU AI Act's risk-based approach, and mitigates business and reputational risks.

#4about 7 minutes

Using MLOps infrastructure to implement model governance

The MLOps lifecycle, including artifact repositories and model registries, provides the technical foundation for proving performance and ensuring reproducibility.

#5about 4 minutes

Differentiating between model interpretability and explainability

Interpretability provides a technical understanding of model behavior for engineers, while explainability communicates decisions to non-technical stakeholders.

#6about 3 minutes

The four core principles of explainable AI

Explanations must be meaningful to the target audience, accurately reflect the model's process, and operate within the model's knowledge limits.

#7about 3 minutes

Applying explainable AI to a recruiting use case

Techniques like anchor explanations and counterfactuals can answer key HR questions about why a candidate was selected and how certain the model is.

#8about 1 minute

Auditing AI systems using MLOps and explainability

Combining MLOps infrastructure for reproducibility with XAI tools enables internal and external auditors to verify model decisions and compliance.

#9about 2 minutes

Conclusion and handling GDPR deletion requests

A discussion on maintaining reproducibility and compliance when faced with GDPR data deletion requests, emphasizing the importance of thorough documentation.