Financial Engineering Seminar Series: Model Risk Management and Governance in AI

close up at the red cube to identify red danger risk cube among other normal incident. Risk management concept for business and personal planning for risk mitigation with information and decision

Abstract

A much-debated question in finance currently is how to manage AI models, since past guidance documents like SR 11-7 seem inadequate to validate, control and maintain them. AI entering finance echoes the 1890s introduction of rapid electric trolleys on streets in my home borough of Brooklyn, where previously they were slow and horse drawn. The famous Brooklyn baseball team was indeed originally called the Trolley Dodgers, reflecting pride at the advances in technology in a modernizing city. But also, the injuries and deaths from electric trolley collisions required better signage, guard rails, and risk management. This talk discusses doing the equivalent for AI models. First, a three-tiered model of AI is proposed to reflect its current use in banking, and corresponding specific global AI regulation. Second, tools are presented to reflect the strong dependences on training data, and to quantify the standard errors of measurement and confidence limits on AI calculations. Finally, governance of LLMs (popular on the buy side) will be explored, with methodologies for ongoing governance monitoring.

Biography

Jonathan is the Founder and CEO of Delta Vega, a New York-based financial engineering (FE) consultancy. An FE from 2000, he has focused on risk management since the Great Financial Crisis. This includes pricing CDOs in the Lehman bankruptcy, validating improved VaR models with EWMA for Morgan Stanley, consent work order for the JP Morgan London Whale episode, and the LIBOR transition at Natixis. A common theme has been doing model risk management (MRM) per Fed SR 11-7, PRA SS1/23 in the UK, and other standards, most recently at Citibank and Jefferies. These are codified in a forthcoming textbook – where Jonathan is the lead author – on MRM and governance, intended for FE masters programs. Jonathan holds a Ph.D. in physics from Berkeley, and an M.A. in mathematics of finance from Columbia. He was a postdoc and research scientist at the Harvard-Smithsonian Center for Astrophysics from 1990 to 2000.

At any time, photography or videography may be occurring on Stevens’ campus. Resulting footage may include the image or likeness of event attendees. Such footage is Stevens’ property and may be used for Stevens’ commercial and/or noncommercial purposes. By registering for and/or attending this event, you consent and waive any claim against Stevens related to such use in any media. See Stevens' Privacy Policy for more information.