Machine Learning Framework for GSE Model Development
To fulfill its mission as a Government Sponsored Entity (GSE), our client conducts business in the U.S. secondary mortgage market and works with a national network of mortgage lending customers.
Challenge
Our client recognized that their current risk governance processes were not efficiently managing access to machine learning technology, validating model risk and providing timely model implementation in support of their internal business customers. The client determined that the formulation of an intake process for machine learning tools and implementation of a parallel risk assessment process would allow for efficient access to machine learning tools, enhance the effective determination of risk and better serve internal customers by reducing model concept to implementation delivery timelines.
The implementation of parallel risk assessment processes and an intake process for machine learning tools involved coordination of multiple departments, definition and automation of new intake processes, and creation of parallel review processes by the first and second line to deliver:
Efficient processes and system to allow appropriate risk governance access to machine learning tools
Parallel validation of training data prior to model implementation
Review of the algorithm design for the machine learning model
Reduction of implementation findings
The early involvement of the second line in the model development process allowed for detection of any issues prior to their final review and reduced the overall model development to implementation timeline. Access to digital technologies allowed the model developers and business areas to utilize the best tools to quickly and efficiently meet their business needs. The new machine learning tools also created efficiencies in the second line’s validation processes as they can now quickly develop challenger models for validation purposes.
The implementation of these processes required a robust change management effort and risk governance expertise to quickly and effectively develop processes which would support a framework to allow the use of new technologies to deliver to the business and modelers to meet their business requirements.
Solution & Delivery
Clarendon Partners transformed the introduction and governance of machine learning tools through the following activities:
Developed processes and an automated system to gain access to machine learning tools
Recommended a new organizational structure for the first line to manage the development and monitoring of machine learning models
Designed processes for the second line of defense to conduct parallel review for machine learning models
Created efficiencies in the documentation creation and review processes for the model developer guide through the use of a machine learning generated guide
Implemented a management governance process to govern business use of AI and RPA to increase their process efficiencies
Recommended risk rating review structure for operational models
Tested new machine learning framework by providing project management for machine learning model development and implementation within prescribed business requirements and deadlines
These initiatives established the methodologies for the organization to meet their internal customer requirements using current technology thereby introducing a framework to govern both operational and financial models.
Project Length: 8 months
Impact & Lasting Value
An intake process for machine learning tools has been automated and will allow for the efficient onboarding of new tools and users. Adoption of parallel validation processes by the second line has led to shorter delivery times and fewer findings prior to model implementation. The client is now able to focus on improving the speed of model changes through the use of machine learning to support the business’s regulatory requirements and strategic vision to gain market share.