Back

Speaker "Josh Bottum" Details Back

 

Topic

Workshop: Kubeflow-based ML Model Risk Management with streamlined SR11-7 compliance

Abstract

The lack of updated compliance reporting for machine learning models frequently blocks profitable models from being deployed. Financial organizations need a process to produce reports and artifacts for governmental regulators, such as those defined in the Federal Reserve’s SR11-7 Guidance on Model Risk Management. In this workshop, attendees will receive a valuable briefing of Model Risk Management’s background, requirements and benefits as well as how AI ethics apply to financial organizations. The second section of the workshop will provide an update on Kubeflow, which is a popular open source framework for machine learning on Kubernetes. The final section will review a demonstration of Fairly’s Model Risk Management solution (Fairly AI). Fairly AI leverages Kubeflow and provides a streamlined user experience for deploying models and producing compliance reports that will satisfy SR-117 requirements. 

Profile

Josh is currently a Kubeflow Community Product Manager and a VP of Community Relations at Arrikto.  Since the Kubeflow v0.5 release (in 2019), Josh has helped to improve the Kubeflow user experience by triaging GitHub issues, running user surveys, gathering requirements, updating roadmaps, writing blog posts, and conducting demonstrations and presentations. Arrikto is a major code contributor to the open-source Kubeflow Community project.