Daniel Abib kicks off this week's AWS roundup with significant moves in data performance and model customization. For data engineers, Amazon Redshift stands out with a 7x performance boost for uncached queries, slashing wait times for dynamic dashboards and ETL pipelines where cache hits are rare.

On the ML front, NVIDIA Nemotron 3 Super has landed on Amazon Bedrock. This addition expands unified API access to high-performance models suited for complex reasoning and code generation. Meanwhile, the new Nova Forge SDK simplifies fine-tuning Amazon Nova models, allowing teams to adapt foundational models to specific domains without managing underlying infrastructure.

Infrastructure reliability also saw upgrades. Amazon EKS now guarantees a 99.99% SLA for Provisioned Control Plane clusters and introduced an 8XL scaling tier, doubling API capacity for heavy workloads like AI training. AWS Lambda added Availability Zone metadata, improving observability for latency-sensitive architectures. Additionally, CloudWatch Logs now accepts HTTP-based ingestion, removing the need for custom agents.

Java developers receive Amazon Corretto 26, the latest LTS release with updated security patches. Beyond core services, AWS opened Kiro AI tools to students and reported 100% agent accuracy using Strands Steering Hooks. With AWS Summits scheduled across Paris, London, and Bengaluru this spring, plus the AWSome Women Summit in Mexico City, the platform continues to push enterprise readiness. Abib, based in São Paulo, notes these updates reflect a broader push toward scalable, observable AI systems.

Source: AWS