Software Architechture AI Day 2026

15:30 - 16:15

Responsible AI Operationalization

What is the presentation about?

This fast-paced, hands-on session turns “ethical AI” from abstract principles into shippable practice. We open with the Anti-Problem for Ethical AI game to surface concrete failure paths (bias, opacity, privacy gaps, audit blind spots). Teams then invert those risks into owned, testable safeguards. We align every safeguard with the OECD AI Principles (e.g., transparency, fairness, accountability, human-centred values), and package results into lightweight Model Cards that document intended use, performance, limits, and monitoring hooks. Attendees leave with a prioritized control set, draft model-card content, and a Responsible AI Ops Framework ready to operationalize Responsible AI in real products.

Prerequisites

Basic familiarity with your use case and data sources.
Awareness of your organization’s risk and compliance touchpoints (privacy, incident response).
A high-level understanding of recognized frameworks helps, but isn’t required: OECD AI Principles.
AI risks to be familiar with (working vocabulary): Reading: https://airisk.mit.edu/

Bias and discrimination, fairness failures (data imbalance, proxy variables, feedback loops).
Opacity, lack of transparency and explainability (hard to trace or justify decisions; weak documentation).
Privacy and data-protection risks.
Security and robustness risks. Adversarial ML, data poisoning, model inversion, membership inference, supply-chain vulnerabilities.
Safety and misuse, hallucinations leading to harmful actions, over-reliance/automation bias, mis/disinformation amplification.
Regulatory risk posture (EU AI Act), prohibited practices, high-risk classifications, transparency duties for certain and general-purpose AI systems.

Learning Objectives

A prioritized list of failure paths (from the Anti-Problem game) inverted into owned, testable safeguards.
A draft Model Card section capturing intended use, limits, subgroup metrics, and monitoring hooks.
A one-page Responsible AI Ops Framework per attendee tagged to OECD Principles for instant policy and legal recognizability.

Speaker

Dr. Larysa Visengeriyeva

Women in Defense Tech

Co-Founder & CEO, Independent Consultant, Author

Why you should attend the AI Day

Architect AI that works in practice: Design scalable, cost-aware, production-ready AI systems.
Bridge software architecture and AI: Combine classic architectures, ML, and GenAI into intelligent systems.
Learn from leading AI architects: Gain real-world insights from experts building AI in production.
Learn online and revisit later: Join live from anywhere and get access to recordings and slides after the conference.