Shipping Gen AI You Can Trust: Building Guarded Applications with Amazon Bedrock Guardrails

Generative AI is powerful, but building AI applications that are safe, reliable, and production-ready requires more than choosing the right model.
In this hands-on session, we’ll explore AI Guardrails and how they can help protect generative AI applications from common risks such as prompt attacks, hallucinations, PII leakage, inappropriate content, and off-limits topics.
We’ll then dive into Amazon Bedrock Guardrails and build a guarded customer-support assistant live. You’ll see how content filters, denied topics, PII redaction, prompt-attack detection, and contextual grounding checks for RAG can be implemented in a practical application.

WHAT YOU’LL LEARN
• Why model-level safety alone isn't enough and how a dedicated guardrail layer works
• The core safeguards available in Amazon Bedrock Guardrails and when to use them
• How to protect both inputs and outputs in Gen AI applications
• How contextual grounding checks can improve reliability in RAG applications
• Practical approaches to making AI safety observable and debuggable
• Working patterns you can apply to your own Gen AI applications

COMMUNITY PARTNER
AiDataYard
Whether you're building your first Gen AI application or already experimenting with RAG and LLM-based systems, this session will give you practical patterns for building AI applications you can trust.
Come learn. Build. Experiment. Ship.