ML Model Risk Management: How to Bring Explainability and Robustness in Production

Finance is one of the most regulated industries in the US, which makes application of cutting-end technologies, like artificial intelligence and machine learning, much harder. Many mission-critical decisions are made based on performance of ML models. That’s why ensuring explainability and robustness on ML models in an adversarial production environment is top priority of any financial organization and introduces a real challenge. 

At the Bugout meetup we hosted Agus Sudjianto, an Executive Vice President and Head of Corporate Model Risk for Wells Fargo, who shared how his team brings explainability and robustness of deployed ML models in production. Agus explained how they make architectural decisions and what models perform the best and deliver transparency in business decision-making.

Agus accumulated his substantial experience in finance from his roles as the modeling and analytics director at Lloyds Banking Group and then the head of Quantitative Risk at Bank of America. He has also shared his expertise with our community. 

You can watch the Bugout YouTube channel on Agus’ talk about how to bring the explainability and robustness of ML models into production here.

You can check out Agus’ tutorials for using deep learning for stochastic differential equations (SDE) here and here.  

If you want to become a part of the fast growing Bugout community of engineers, data scientists, ML practitioners, join our Slack here. Our goal is to support our community members: either by providing technical advice, pair programming sessions, or helping finding a job/hire team members or bringing exposure to projects and articles of our community members. Join the Bugout community today!