AI news for builders and product teamsUpdated Nov 17, 2026, 13:01 UTC
AWS Machine Learning Blog
First-party releases and research from AWS Machine Learning Blog. Headlines and excerpts link to the original articles.
Latest stories
Newest first
AWS published best-practice guidance for administering Amazon SageMaker HyperPod through SageMaker Unified Studio, covering four control layers: organization, project, cluster, and workload. The post advises centralizing cluster and accelerator capacity in one capacity account, separating authorization from scheduling policy, and reviewing project-to-cluster connections across their lifecycle.

AWS now lets users create and manage Amazon SageMaker Spaces on HyperPod EKS clusters directly from the SageMaker Studio UI, including a new IDE and Notebooks tab for launching JupyterLab and Code Editor environments in the browser. Administrators must install the Spaces add-on and configure EKS access entries and per-user identity propagation first.

AWS published a guide for building a voice travel concierge on Amazon Bedrock AgentCore, using Nova 2.5 Sonic for speech-to-speech and Bedrock Knowledge Bases for policy answers. The sample deploys via AWS CDK and connects to a synthetic airline backend through MCP.

Z.ai's GLM 5.3, a 753B-parameter mixture-of-experts model optimized for coding and long-horizon agentic tasks, is now available on Amazon Bedrock for eligible enterprise customers. Z.ai reported a CyberGym benchmark score of 84.5 at release.

AWS published guidance on running Anthropic's Claude Code agentic coding tool with Claude Opus 5.5, Claude Sonnet 5.5, and Claude Sonnet 5 on Amazon Bedrock in AWS GovCloud (US) Regions for regulated workloads. It covers prerequisites, two endpoint surfaces, installation, configuration, and enterprise deployment considerations.

Amazon SageMaker AI introduced the aws-ai-ml skill, available through the Agent Toolkit for AWS, which gives coding agents such as Kiro, Claude Code, and Codex inference optimization and benchmarking capabilities. The skill works with MCP-compatible agents and can be installed locally or used in SageMaker Studio.

AWS published a sample Model Context Protocol server, the Quick Resource Migrator, hosted on Amazon Bedrock AgentCore, that automates cross-account promotion of Amazon Quick agents, action connectors, knowledge bases, flows, and spaces. It is idempotent, copies permissions, and writes versioned Amazon S3 backups before updates.

Amazon Bedrock Managed Knowledge Bases now supports agentic retrieval, exposed through the langchain-aws package, which decomposes multi-part questions into sub-queries and iterates until it judges the evidence sufficient. It is available only on managed knowledge bases, requires boto3 1.43.32 or later, and supports guardrails in BLOCK mode only.

An AWS Machine Learning Blog post explains how to downgrade user roles in Amazon Quick, since the console and update-user API reject direct downgrades from Author to Reader. It details manual deletion-and-recreation and an AWS CLI step-down method, plus asset ownership transfer and post-downgrade reviews.

The AWS Machine Learning Blog describes a reference implementation for building and evaluating a multi-agent supply chain decisioning system using Amazon Bedrock AgentCore Evaluations, Strands Agents SDK, and related AWS services. It outlines built-in, custom, and explainability evaluators applied to an orchestrator and four sub-agents, plus deployment instructions.

An AWS Machine Learning Blog post introduces the Adjudicated Query pattern, which pairs Amazon Quick's chat interface with a deterministic rules engine so compliance officers can sweep thousands of leases while keeping pass/fail decisions outside the model. It includes a reference architecture and a deployable GitHub sample using Amazon Aurora, Lambda, API Gateway, Cognito, and Bedrock.

An AWS Machine Learning Blog post walks through connecting Claude Desktop on Amazon Bedrock to Web Search through Amazon Bedrock AgentCore Gateway, using JWT-based inbound authentication. The setup federates AWS IAM Identity Center with Amazon Cognito as the JWT issuer, and Web Search is available in three AWS Regions.

AWS describes fine-tuning a search agent with multi-turn reinforcement learning (MTRL) on Amazon SageMaker AI, using Qwen3.6-27B with nDCG@10 as reward. The fine-tuned model improved nDCG@10 on three of four held-out benchmarks, with the largest gains on BrowseComp-Plus (+23.7%) and WixQA (+18.4%), and cut BrowseComp-Plus failure rate from 22.89% to 0.68%.

An AWS Machine Learning Blog post describes a four-agent pattern built with the Strands Agents SDK on Amazon Bedrock AgentCore for enterprise cloud migrations. The post reports that on a 300+ application program, the pattern cut infrastructure as code development from 3 to 4 weeks per application to minutes, based on internal project tracking data.

Amazon Quick has launched Live Data in Apps, letting AI-built Quick apps query governed Quick Sight datasets in real time instead of relying on static build-time snapshots. Queries execute as the viewer, so existing row-level and column-level security rules apply, with per-dataset consent and server-side enforcement.

An AWS Machine Learning Blog post shows how to implement Amazon S3 Vectors as a custom persistent memory provider for the NVIDIA NeMo Agent Toolkit (NAT), tested with NAT version 1.6. The walkthrough covers creating S3 Vectors infrastructure, writing a MemoryEditor plugin, configuring an agent workflow, and deploying on Amazon EKS.

Amazon Payments used a multi-objective contextual multi-armed bandit on Amazon SageMaker AI to personalize content across a product acquisition funnel. In a seven-week online A/B test, one customer population saw a high single-digit percentage relative lift in final-funnel conversion, while another saw no improvement.

An AWS Machine Learning Blog post details how to build ambient agents using Amazon Bedrock AgentCore, where event streams like Amazon S3 uploads trigger agent jobs that pause for human input via a single ask_human tool. The reference sample combines AgentCore Runtime, Lambda, DynamoDB, SQS, and a React frontend into a serverless platform.

An AWS Machine Learning Blog post details a step-by-step guide for giving three environment types access to Claude Platform on AWS inference: AWS workload accounts via cross-account SigV4, developer laptops via a workspace-scoped API key, and external workloads via OIDC federation, all sharing one subscription with workspace isolation.

Amazon Quick Sight announced the hierarchy filter, a single control that lets dashboard authors present multi-level drill-down filtering instead of separate controls. It supports up to five levels, dimension fields only, and searches limited to the highest hierarchy level.