MS-EE-L Archives

November 2021

MS-EE-L@LISTSERV.GMU.EDU

Options: Use Proportional Font
Show HTML Part by Default
Condense Mail Headers

Message: [<< First] [< Prev] [Next >] [Last >>]
Topic: [<< First] [< Prev] [Next >] [Last >>]
Author: [<< First] [< Prev] [Next >] [Last >>]

Print Reply
Content-Type:
multipart/alternative; boundary="_000_BL0PR05MB5107BFE64298553A48B36AEAA6959BL0PR05MB5107namp_"
Date:
Fri, 12 Nov 2021 14:37:03 +0000
Reply-To:
Patricia M Sahs <[log in to unmask]>
Subject:
From:
Patricia M Sahs <[log in to unmask]>
Message-ID:
MIME-Version:
1.0
Sender:
Comments:
Parts/Attachments:
text/plain (2975 bytes) , text/html (8 kB)
ECE Distinguished Seminar Series



Delivering Enterprise-AI: Challenges, Learnings and Opportunities



Dr. Rajarshi Das

Co-founder & Chief-Scientist

FatBrain.ai



November 12, 10:00am

Joining in-person at ENGR 4201

Joining via Zoom: https://gmu.zoom.us/j/97539840707



To attend in-person, please RSVP by Nov. 11 via this link<https://forms.gle/xSqo6WXjpB5gfGFy6>.

All attendees must complete Mason COVID Health✓™<https://www2.gmu.edu/mason-covid-health-check> and receive a "green light" status on the day of the event.



Abstract



While AI-driven computational systems like AlphaGo and AlphaFold has attracted wide media attention in achieving super-human level of performance in their target domain, reports of successful demonstrations of Enterprise AI, i.e., general AI-automation engines at enterprise scale and complexity, has been lacking.   This lecture reports on the deployment of a general, scalable automated decision-engine configured to fight financial crime at a leading global bank that outperformed the effectiveness and efficiency of the bank’s own incumbent system comprising 100 complex siloed risk models and 1000 human experts.    The success of our approach hinged critically in being able to automatically, scalably and dynamically score risk at different levels of hierarchies within the bank and offering a practical yet principled way to integrate and manage risk within and across these levels in a transparent fashion.



To fight financial crime and comply with the 2001 USA Patriot Act, modern global banks may process up 10B transactions per day where each transaction can trigger any one of 100s of siloed models and hard-coded rules.  This results in a deluge of potentially false positive compliance alerts which are then threshold-controlled to manage the volume of caseload requiring human investigative effort and related compliance paperwork.  Constrained to a finite number of human experts, banks are thus exposed to uncontrolled compliancy risk at the institutional level. We enabled the bank to advance to a dynamic behavior risk framework that proactively and automatically learns, scores and promotes new cases, risks, and outcomes in real time. Our AI-automation software system, deployed in one quarter within the bank’s regulated infrastructure, tollgate practices, and secure IT policies, helped the bank to modernize with a unified model consolidation framework that was compatible with the bank’s existing processes and IT framework in a “plug and play” fashion.



This lecture retraces some of the key challenges and learnings in delivering AI-automation software to enterprises and outlines potential opportunities ahead.  It is our belief that every business, including the vast majority of small businesses, will need AI-automation to help them think, build and grow their businesses.  Indeed, our mission is to deliver Enterprise AI for All.




ATOM RSS1 RSS2