Challenges with model governance
Businesses implement ML models across their entire organization, spanning a large spectrum of usecases. When you start deploying more than just a couple models in production, a lot of complex governance and management challenges arise.
Almost all the governance needs for ML are associated with data and are tied directly to the data management practice in your organization. For example, what data can be used for certain applications, who should be able to access what data, and based on what data are models created.
Some of the other unique governance challenges that you could encounter are:
How to gain visibility into the impact your models have on your customers?
How can you ensure you are still compliant with both governmental and internal regulations?
How does your organization’s security practices apply to the models in production?
Ultimately, the needs for ML governance can be distilled into the following key areas: model visibility, and model explainability, interpretability, and reproducibility.