Guardrails in Action - Ensuring Safe AI with Azure AI Content Safety by Udaiappa Ramachandran

In this session, Udaiappa Ramachandran (Udai), CTO/CSO at Akumina Inc. and Microsoft MVP (AI), walks through how to build safe, trustworthy, and compliant AI systems using Azure AI Content Safety, Guardrails, and the Responsible AI Toolkit. You’ll learn why safe AI matters, how harm detection works, and how to protect users and businesses from biased, offensive, or unsafe outputs. Udai also demonstrates real-world scenarios across various sectors, including education, social media, gaming, e-commerce, news moderation, and LLM-powered applications. This presentation covers: Why safe AI is essential for fairness, inclusivity, and enterprise trust Azure AI Content Safety capabilities (text, image, severity scores, harm categories) Prompt Shields, Groundedness Detection & Protected Material Detection Understanding harm categories: Hate, Sexual, Violence, Self-Harm Severity scoring (0–7) and multi-label classification Guardrails, blocklists, and organization-specific safety policies Real-world use cases for AI moderation How to build a safe AI chat workflow in Azure AI Foundry Key takeaways for building responsible, reliable, and ethical AI solutions Whether you're a developer, architect, product leader, or AI enthusiast, this session will help you confidently implement content safety and responsible AI protections in modern AI apps.