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.

From learning to earning

Jobs that call for the skills explored in this talk.