Day 2 kicked off with Matt Garman’s keynote, and he opened with a quote that instantly stood out as my favorite of the day
“Is it possible? …..Why not! “
There was a flurry of announcements, as expected, and a few favorites stand out below.
Amazon Nova Forge: Build your own Frontier Models using Nova
Applications that depend on deep domain expertise or specialized business contexts require models that can accurately grasp their proprietary knowledge, workflows, and distinct needs. Approaches and strategies like prompt engineering, RAG, fine-tuning, continued pre-training has its own limitations and its difficult to steer the model to specific domains of interest.
This is where Nova Forge introduces its unique “open training” approach, providing exclusive access to pre-trained, mid-trained, and post-trained Nova model checkpoints, enabling customers to combine their proprietary data with Amazon Nova-curated datasets at every stage of model development. Nova Forge enables organizations to create their own optimized versions of Nova, known as “Novellas,” by integrating their proprietary data with Nova’s advanced capabilities.
Nova Forge allows you to build models that deeply understand your organization’s proprietary knowledge and reflects your expertise, while preserving general capabilities like reasoning and minimizing risks like forgetting.
Frontier agents for Operational excellence,Security and Development
Purpose-built frontier agents are the way forward, and there are now three of them in the toolkit.
AWS Security Agent helps protect applications across the entire development lifecycle. It performs automated security reviews based on your organization’s policies and runs context-aware penetration tests to catch vulnerabilities early, from design through deployment.
Security teams set their standards once in the Security Agent console, including requirements for encryption, authentication, and logging. The agent then continuously checks architecture documents and code for compliance, offering clear remediation guidance when issues appear.
For deployment, teams can configure penetration testing scopes, and Security Agent automatically builds application context, runs attack simulations, and identifies vulnerabilities. This ensures consistent enforcement of security policies, scales security reviews to keep up with development speed, and turns penetration testing into an on-demand, continuous process that reduces risk exposure.
AWS DevOps Agent helps prevent and resolve incidents while continuously enhancing application reliability and performance across AWS, multicloud, and hybrid environments. It investigates issues and pinpoints operational improvements much like an experienced DevOps engineer—by understanding your resources and their relationships, integrating with observability tools, runbooks, code repositories, and CI/CD pipelines, and correlating telemetry, code, and deployment data to uncover how components interact.
And finally, Kiro Autonomous Agent independently handles development tasks without needing constant human supervision, so you can stay focused on higher-value work. In an agentic development workflow, it can take a high-level task description, create an implementation plan, write code across multiple repositories, run tests, and open pull requests in the background. The agent works asynchronously, maintaining progress even when you are not in an active session.
Amazon Nova 2 Omni
Amazon Nova 2 Omni is an all-in-one multimodal reasoning and image generation model that supports text, image, video, and speech inputs while producing both text and image outputs, making it one of the few models in the industry with this breadth of capability. It enables multimodal understanding, natural language–driven image generation and editing, and robust speech transcription in a single unified model.?
Unlike traditional setups where teams must stitch together multiple specialized models for different input and output types, Nova 2 Omni lets organizations handle these workloads with one model, reducing integration complexity and operational overhead. This unified approach helps accelerate application development and lowers costs while supporting use cases such as marketing content creation, customer support call transcription, video analysis, and documentation with visual aids.
Optimize CPU for RDS SQL Server & Developer Edition Support
As a database enthusiast, this is something that I have been waiting for year. Pretty sure, this is one of the most requested features from our customers as well.
Amazon RDS for SQL Server launches optimize CPU with support for M7i and R7i instance families, which reduce prices by up to 55% compared to equivalent previous generation instances. Optimize CPU optimizes Simultaneous Multi-threading (SMT) configuration to reduce commercial software charges. Optimize CPU disables SMT for instances with 2 or more physical CPU cores. This reduces the number of vCPUs, and the corresponding commercial software charges by 50% while providing the same number of physical CPU cores, and near equivalent performance.
RDS SQL Server now offers Microsoft SQL Server 2022 Developer Edition. SQL Server Developer Edition is a free edition of SQL Server that contains all the features of Enterprise Edition and can be used in any non-production environment. This enables customers to build, test, and demonstrate applications using SQL Server while reducing costs and maintaining consistency with their production database configurations. Previously, customers that created Amazon RDS for SQL Server instances for development and test environments had to use SQL Server Standard Edition or SQL Server Enterprise Edition, which resulted in additional database licensing costs for non-production usage.
Amazon Bedrock AgentCore now includes Policy & Evaluations
Policy in AgentCore works with AgentCore Gateway to intercept every tool call in real time, keeping agents within defined policies without adding noticeable latency. Teams can define rules in natural language, which are automatically translated into Cedar, the AWS open-source policy language, so development, compliance, and security stakeholders can author, understand, and audit controls without custom code.
AgentCore Evaluations lets developers test and continuously monitor agent behavior against real-world scenarios to improve quality and catch issues before they affect many customers. It includes 13 built-in evaluators for dimensions like helpfulness, tool choice, and accuracy, and also supports custom model-based scoring, which greatly reduces the effort needed to build and maintain evaluation infrastructure.
While these did not make it into my top five, they are still important announcements for our customers and deserve a special mention.
Database Savings Plans with up to 35% savings
S3 Vectors GA
Bedrock adds 18 fully managed open-weight models
Mistral Large 3 and Ministral 3 family now available on Bedrock
Amazon S3 Batch Operations improvements
Lambda Durable Functions
Thanks for reading, and stay tuned for more updates here! #LetsGoBuild
