BEGIN:VCALENDAR VERSION:2.0 PRODID:IEEE vTools.Events//EN CALSCALE:GREGORIAN BEGIN:VTIMEZONE TZID:Asia/Kolkata BEGIN:STANDARD DTSTART:19451014T230000 TZOFFSETFROM:+0630 TZOFFSETTO:+0530 TZNAME:IST END:STANDARD END:VTIMEZONE BEGIN:VEVENT DTSTAMP:20260902T174251Z UID:6D4579DD-607F-403F-982D-C826E8F678ED DTSTART;TZID=Asia/Kolkata:20260911T190000 DTEND;TZID=Asia/Kolkata:20260911T200000 DESCRIPTION:Artificial intelligence is changing not only software capabilit ies but also the way systems must be specified\, built\, tested\, released and operated. An AI-enabled application combines models\, prompts\, conte xtual data\, retrieval mechanisms\, tools\, integrations and human decisio ns. Its behaviour may change when any of these components changes. Traditi onal software-development controls must therefore be extended to address u ncertainty\, data provenance\, security\, accountability and continuous ev aluation.\nThis technical talk presents an evidence-driven AI System Devel opment Life Cycle comprising five stages: Intent\, Build\, Review\, Deploy and Verify. It explains how to define a bounded use case and prohibited a ctions\; design traceable AI-system architecture\; manage model\, prompt\, data and tool versions\; construct representative evaluation sets\; and a ssess outcome quality\, security\, privacy\, latency and cost. Particular attention will be given to prompt injection\, data leakage\, unsafe tool u se\, human oversight\, release authorization\, rollback and incident escal ation.\nThe session will also examine the evidence required at each lifecy cle gate\, the conditions that should prevent an AI system from entering p roduction\, and the monitoring needed after deployment. An illustrative wo rkflow will show how engineering\, security\, governance and business team s can work within a common lifecycle without diluting accountability. Part icipants will leave with a practical framework for moving from an AI exper iment to a controlled\, measurable and trustworthy production system.\n\nS peaker(s): Dr. Ashwini Kumar Rath\, \n\nVirtual: https://events.vtools.iee e.org/m/575480 LOCATION:Virtual: https://events.vtools.ieee.org/m/575480 ORGANIZER:amiya87@gmail.com SEQUENCE:22 SUMMARY:Engineering Trustworthy AI: An Evidence-Driven SDLC for AI-Enabled Systems URL;VALUE=URI:https://events.vtools.ieee.org/m/575480 X-ALT-DESC:Description: <br /><p>Artificial intelligence is changing not on ly software capabilities but also the way systems must be specified\, buil t\, tested\, released and operated. An AI-enabled application combines mod els\, prompts\, contextual data\, retrieval mechanisms\, tools\, integrati ons and human decisions. Its behaviour may change when any of these compon ents changes. Traditional software-development controls must therefore be extended to address uncertainty\, data provenance\, security\, accountabil ity and continuous evaluation.<br>This technical talk presents an evidence -driven AI System Development Life Cycle comprising five stages: Intent\, Build\, Review\, Deploy and Verify. It explains how to define a bounded us e case and prohibited actions\; design traceable AI-system architecture\; manage model\, prompt\, data and tool versions\; construct representative evaluation sets\; and assess outcome quality\, security\, privacy\, latenc y and cost. Particular attention will be given to prompt injection\, data leakage\, unsafe tool use\, human oversight\, release authorization\, roll back and incident escalation.<br>The session will also examine the evidenc e required at each lifecycle gate\, the conditions that should prevent an AI system from entering production\, and the monitoring needed after deplo yment. An illustrative workflow will show how engineering\, security\, gov ernance and business teams can work within a common lifecycle without dilu ting accountability. Participants will leave with a practical framework fo r moving from an AI experiment to a controlled\, measurable and trustworth y production system.</p> END:VEVENT END:VCALENDAR