AWS Architecture Blog

Resilience testing your agents on Amazon Bedrock AgentCore

Resilience testing your agents on Amazon Bedrock AgentCore

When an AI agent’s tools fail or return corrupted data, does it fail safe or confidently fabricate an answer? Learn how to inject tool-level faults (cancelled calls, corrupted responses, stripped fields) into a Strands agent on Amazon Bedrock AgentCore using AWS Fault Injection Service, and build the observability to tell a safe failure from a silent one.

How Rakbank built centralized observability at scale with Amazon OpenSearch Service

How Rakbank built centralized observability at scale with Amazon OpenSearch Service

Rakbank re-platformed logging for its retail digital banking workloads on a cross-account, Multi-AZ Amazon OpenSearch Service design that ingests ~150 GB/day through Fluent Bit, using hot/UltraWarm/cold tiering and Amazon S3 archival for 7-year compliance retention and ~55% cost savings over an all-hot cluster.

Scale distributed AI workloads with Ray and Anyscale on Amazon EC2

Scale distributed AI workloads with Ray and Anyscale on Amazon EC2

Building multimodal AI/ML pipelines on Amazon EC2 means scaling distributed compute, controlling costs across stages, and keeping up with new models and hardware. This post shows how the Ray-powered Anyscale application on AWS helps you deploy Ray across three compute options and scale from prototype to production without code changes.

Inside an enterprise AI agent: A governed, observable loop architecture

Inside an enterprise AI agent: A governed, observable loop architecture

This post lays out a five-layer reference architecture for governed, observable enterprise AI agents on AWS: an API gateway, a five-phase agent runtime, a knowledge and state tier, platform services, and a cross-cutting trust and control framework. Each layer maps to managed AWS services, with the design patterns worth copying and the mistakes to avoid.

Self-service regional failover with Amazon Application Recovery Controller Region Switch and centralized ingress controls

Self-service regional failover with Amazon Application Recovery Controller Region Switch and centralized ingress controls

Application teams in regulated enterprises can execute self-service regional failover with Amazon Application Recovery Controller (ARC) Region Switch, even behind centrally managed ingress proxy fleets. Learn how Transit Gateway peering, Route 53 Private Hosted Zones, and VPC Lattice preserve security controls while removing cross-team coordination during DR.