BEGIN:VCALENDAR VERSION:2.0 PRODID:IEEE vTools.Events//EN CALSCALE:GREGORIAN BEGIN:VTIMEZONE TZID:America/New_York BEGIN:DAYLIGHT DTSTART:20260308T030000 TZOFFSETFROM:-0500 TZOFFSETTO:-0400 RRULE:FREQ=YEARLY;BYDAY=2SU;BYMONTH=3 TZNAME:EDT END:DAYLIGHT BEGIN:STANDARD DTSTART:20261101T010000 TZOFFSETFROM:-0400 TZOFFSETTO:-0500 RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=11 TZNAME:EST END:STANDARD END:VTIMEZONE BEGIN:VEVENT DTSTAMP:20260912T193719Z UID:C87EA8F4-71E6-4CCE-9980-653C94071B50 DTSTART;TZID=America/New_York:20260917T190000 DTEND;TZID=America/New_York:20260917T200000 DESCRIPTION:Generative AI works beautifully in the demo. The harder questio n is what happens once real customers arrive.\n\nThis talk comes out of my experience architecting generative AI for large North American financial institutions. I walk four ways these systems surprise the teams that deplo y them: what they inherit from the systems underneath them\, what they cos t once real customers use them\, how they can be attacked through the text they read\, and how they can leak one session’s data into another sessi on. The fix is the same move each time. Keep the AI\, and move the guarant ee outside it: a clear source of truth for what it inherits\, a spending l imit with an owner for what it costs\, a gatekeeper the model cannot talk past for what attacks it\, and a boundary enforced on every step against l eaks. The few AI terms the talk needs are introduced in plain language up front\, so no background in AI is required.\n\nThis session is for enginee rs\, architects\, and technology leaders putting AI into real systems. It is a practitioner's account of what breaks in production and the engineeri ng that holds it together\, not a framework overview.\n\nSpeaker(s): Rohi th Rayapu \n\nAgenda: \n7:00PM - Introduction of IEEE Hamilton Section\n\n 7:15PM - Presentation\n\n8:00PM - Q&A\n\n8:15PM - Refreshments\n\nRoom: Bo ardroom\, Bldg: Queen Elizabeth Park Community and Cultural Centre\, 2302 Bridge Rd\, \, Oakville\,\, Ontario\, Canada\, L6L 2G6 LOCATION:Room: Boardroom\, Bldg: Queen Elizabeth Park Community and Cultura l Centre\, 2302 Bridge Rd\, \, Oakville\,\, Ontario\, Canada\, L6L 2G6 ORGANIZER:sneh@rchilli.com, eduardo.gomez.hennig@ieee.org SEQUENCE:19 SUMMARY:Production AI in Financial Services: What Breaks\, and the Controls That Hold It Together URL;VALUE=URI:https://events.vtools.ieee.org/m/572559 X-ALT-DESC:Description: <br /><p>Generative AI works beautifully in the dem o. The harder question is what happens once real customers arrive. &nbsp\; <br><br>This talk comes out of my experience architecting generative AI fo r large North American financial institutions. I walk four ways these syst ems surprise the teams that deploy them: what they inherit from the system s underneath them\, what they cost once real customers use them\, how they can be attacked through the text they read\, and how they can leak one se ssion&rsquo\;s data into another session. The fix is the same move each ti me. Keep the AI\, and move the guarantee outside it: a clear source of tru th for what it inherits\, a spending limit with an owner for what it costs \, a gatekeeper the model cannot talk past for what attacks it\, and a bou ndary enforced on every step against leaks. The few AI terms the talk need s are introduced in plain language up front\, so no background in AI is re quired. <br><br>This session is for engineers\, architects\, and technolog y leaders putting AI into real systems. It is a practitioner's account of what breaks in production and the engineering that holds it together\, not a framework overview.&nbsp\;</p><br /><br />Agenda: <br /><p>7:00PM - Int roduction of IEEE Hamilton Section</p>\n<p>7:15PM - Presentation</p>\n<p>8 :00PM - Q&amp\;A</p>\n<p>8:15PM - Refreshments</p> END:VEVENT END:VCALENDAR