Amazon SageMaker Unified Studio now provides real-time notifications for data catalog activities, enabling data teams to stay informed of subscription requests, dataset updates, and access approvals. With this launch, customers receive real-time notifications for catalog events including new dataset publications, metadata changes, and access approvals directly within the SageMaker Unified Studio notification center. This launch streamlines collaboration by keeping teams updated as datasets are published or modified. The new notification experience in SageMaker Unified Studio is accessible from a “bell” icon in the top right corner of the project home page. From here, you can access a short list of recent notifications including subscription requests, updates, comments, and system events. To see the full list of all notifications, you can click on “notification center” to see all notifications in a tabular view that can be filtered based on your preferences for data catalogs, projects and event types. Notifications within SageMaker Unified Studio is available in all https://docs.aws.amazon.com/sagemaker-unified-studio/latest/adminguide/supported-regions.html. To learn more, refer to the https://docs.aws.amazon.com/sagemaker-unified-studio/latest/userguide/sagemaker-events-and-notifications.html

Amazon SageMaker AI Projects now supports provisioning custom machine learning (ML) project templates from Amazon S3. Administrators can now manage ML templates in SageMaker AI studio so data scientists can create standardized ML projects to meet their organizational needs. Data scientists can use Amazon SageMaker AI Projects to create standardized ML projects that meet organizational requirements and automate ML development workflows. Administrators define standardized ML project templates that include end-to-end development patterns. By provisioning custom templates from Amazon S3, administrators can define standardized project templates and provide access to these templates directly in the SageMaker AI studio for data scientists, ensuring all ML projects follow organizational standards. SageMaker AI Projects custom template S3 provisioning is available in all AWS Regions where SageMaker AI Projects is available. To learn more, visit SageMaker AI Projects documentation, and SageMaker AI Studio.

Unifying data silos, Amazon SageMaker Lakehouse seamlessly integrates S3 data lakes and Redshift warehouses, enabling unified analytics and AI/ML on a single data copy through open Apache Iceberg APIs and fine-grained access controls.

The next generation of Amazon SageMaker is now available in AWS Europe (Stockholm). Amazon SageMaker is the center for all your data, analytics, and AI. From SageMaker Unified Studio, you can discover your data and put it to work using familiar AWS tools for model development, generative AI app development, data processing, and SQL analytics. Unified access to data is provided by Amazon SageMaker Lakehouse, and catalog and governance features are available via SageMaker Catalog (built on Amazon DataZone) to help you meet enterprise security requirements. For more information on AWS Regions where the next generation of Amazon SageMaker is available, see https://docs.aws.amazon.com/sagemaker-unified-studio/latest/adminguide/supported-regions.html. To get started, see the following resources: https://aws.amazon.com/sagemaker/ https://docs.aws.amazon.com/sagemaker/

Amazon SageMaker Catalog integrates with Amazon S3 Tables, making it easy to discover, share, and govern S3 Tables for users to access and query the data with all Apache Iceberg–compatible tools and engines. With Amazon SageMaker Catalog, built on Amazon DataZone, users can securely discover and access approved data and models using semantic search with generative AI–created metadata, or just ask Amazon Q Developer with natural language to find your data. S3 Tables deliver the first cloud object store with built-in Apache Iceberg support. Data publishers can onboard S3 tables to SageMaker Lakehouse and enhance their discoverability by adding them to the SageMaker Catalog. Publishers have the flexibility to either directly publish tables or enrich them with valuable business metadata, making it easier for all users to understand and find the data they need. On the consumption side, users can search for relevant tables, request access through a subscription workflow (subject to publisher approval), and leverage this data for advanced analytics and AI development projects. This end-to-end workflow significantly improves data accessibility, governance, and utilization of S3 Tables across the organization. SageMaker Catalog with S3 Tables support is available in all AWS Regions where Amazon SageMaker is available. To learn more, visit Amazon SageMaker. Get started with S3 Tables and publish using user documentation.

Today, Amazon SageMaker Unified Studio announced new capabilities allowing SageMaker projects to add custom tags to resources created through the project. This helps customers enforce tagging standards that conform to Service Control Policies (SCP) and helps enable cost tracking reporting practices on resources created across the organization. As an Amazon SageMaker Unified Studio administrator, you can configure a project profile with tag configurations that will be pushed down to all projects using the project profile. Project profiles can be setup to pass Key and Value tag pairings or pass the Key of the tag with a default Value that can be modified during project creation. All tag values passed to the project will result in the resources created by that project being tagged. This provides administrators a governance mechanism that enforces project resources have the expected tags. This first release of custom tags for project resources is supported only through application programming interface (API). Custom tags for project resources capability is available in all https://aws.amazon.com/about-aws/global-infrastructure/regional-product-services/ where Amazon SageMaker Unified Studio is supported, including: Asia Pacific (Tokyo), Europe (Ireland), US East (N. Virginia), US East (Ohio), US West (Oregon), Europe (Frankfurt), South America (São Paulo), Asia Pacific (Seoul), Europe (London), Asia Pacific (Singapore), Asia Pacific (Sydney), Canada (Central), Asia Pacific (Mumbai), Europe (Paris), Europe (Stockholm) To learn more, visit https://aws.amazon.com/sagemaker/ then get started with the https://docs.aws.amazon.com/datazone/latest/APIReference/Welcome.html.

Today, we are announcing improvements to the Amazon Q Developer chat experience in Amazon SageMaker Unified Studio Jupyter notebooks and adding https://docs.aws.amazon.com/amazonq/latest/qdeveloper-ug/command-line.html in Jupyter notebooks and Code Editor. By integrating with Model Context Protocol (MCP) servers, Amazon Q Developer is aware of your SageMaker Unified Studio https://docs.aws.amazon.com/sagemaker-unified-studio/latest/userguide/projects.html resources, including data, compute, and code, and provides personalized assistance for data engineering and machine learning development work. These new capabilities provide highly relevant responses to assist with tasks like code refactoring, file modification, and troubleshooting. This helps data scientists and data engineers quickly set up their integrated development environments and work more efficiently while maintaining transparency into how the AI assistant is acting on their behalf. These features are available at no additional cost with the Amazon Q Developer Free Tier in all https://docs.aws.amazon.com/sagemaker-unified-studio/latest/adminguide/supported-regions.html where Amazon SageMaker Unified Studio is available. To make even more use of these features, we recommend enabling Amazon Q Developer Pro. To do so, please refer to the https://docs.aws.amazon.com/sagemaker-unified-studio/latest/adminguide/amazonq.html.

Enable priority-based resource allocation, fair-share utilization, and automated task preemption for optimal compute utilization across teams.

Today, Amazon SageMaker introduces new multi-adapter inference capabilities that unlock exciting possibilities for customers using pre-trained language models. This feature allows you to deploy hundreds of fine-tuned LoRA (Low-Rank Adaptation) model adapters behind a single endpoint, dynamically loading the appropriate adapters in milliseconds based on the request. This enables you to efficiently host many specialized LoRA adapters built on a common base model, delivering high throughput and cost-savings compared to deploying separate models. With multi-adapter inference, you can quickly customize pre-trained models to meet diverse business needs. For example, marketing and SaaS companies can personalize AI/ML applications using each customer's unique images, communication style, and documents to generate tailored content in seconds. Similarly, enterprises in industries like healthcare and financial services can reuse a common LoRA-powered base model to tackle a variety of specialized tasks, from medical diagnosis to fraud detection, by simply swapping in the appropriate fine-tuned adapter. This flexibility and efficiency unlocks new opportunities to deploy powerful, adaptable AI across your organization. The multi-adapter inference feature is generally available in: Asia Pacific (Tokyo, Seoul, Mumbai, Singapore, Sydney, Jakarta), Canada (Central), Europe (Frankfurt, Stockholm, Ireland, London), Middle East (UAE), South America (Sao Paulo), US East (N. Virginia, Ohio), and US West (Oregon). To get started, refer to the https://docs.aws.amazon.com/sagemaker/latest/dg/realtime-endpoints-adapt.html for information on using LoRA and managing model adapters.  

この記事はUMITRONAdventCalendar日目の記事ですEnglishpage目次まえがき大まかな流れAmazonSageMakerGroundTru

Amazon SageMaker now supports direct connectivity to Oracle, Amazon DocumentDB, and Microsoft SQL Server databases, expanding the available data integration capabilities in Amazon SageMaker Lakehouse. This enhancement enables customers to seamlessly access and analyze data from these databases. With these new data source connections, customers can directly query data and build ETL flows from their Oracle, Amazon DocumentDB, and Microsoft SQL Server databases. This integration simplifies data and AI/ML workflows by allowing you to work with your data alongside AWS data, analytics and AI capabilities. Support for these new data sources is available in all AWS Regions where Amazon SageMaker Unified Studio is available. For the most up-to-date information about regional availability, visit the AWS Region table. To learn more about connecting to data sources in Amazon SageMaker Lakehouse, visit the documentation.

Visual ETL in Amazon SageMaker now offers 9 new built-in transforms: “Derived column”, “Flatten”, “Add current timestamp”, “Explode array or map into rows”, “To timestamp”, “Array to columns”, “Intersect”, “Limit” and “Concatenate columns”. Visual ETL in Amazon SageMaker provides a drag-and-drop interface for building ETL flows and authoring flows with Amazon Q Developer. With these new transforms, ETL developers can quickly build more sophisticated data pipelines without having to write custom code for common transform tasks. Each of these new transforms address a unique data processing need. For example, use “Derived column” to define a new column based on a math formula or SQL expression, use “To timestamp” to convert a column to timestamp type, or build a new string column using the values of other columns with an optional spacer with the “Concatenate columns” transform. This new feature is now available in all AWS regions where Amazon SageMaker is available. Access the supported https://docs.aws.amazon.com/sagemaker-unified-studio/latest/adminguide/supported-regions.html for the most up-to-date availability information. To learn more, visit our Amazon SageMaker https://docs.aws.amazon.com/sagemaker-unified-studio/latest/userguide/visual-etl-supported-transforms.html.

Today, we are introducing the new ModelTrainer class and enhancing the ModelBuilder class in the SageMaker Python SDK. These updates streamline training workflows and simplify inference deployments. The ModelTrainer class enables customers to easily set up and customize distributed training strategies on Amazon SageMaker. This new feature accelerates model training times, optimizes resource utilization, and reduces costs through efficient parallel processing. Customers can smoothly transition their custom entry points and containers from a local environment to SageMaker, eliminating the need to manage infrastructure. ModelTrainer simplifies configuration by reducing parameters to just a few core variables and providing user-friendly classes for intuitive SageMaker service interactions. Additionally, with the enhanced ModelBuilder class, customers can now easily deploy HuggingFace models, switch between developing in local environment to SageMaker, and customize their inference using their pre- and post-processing scripts. Importantly, customers can now pass the trained model artifacts from ModelTrainer class easily to ModelBuilder class, enabling a seamlessly transition from training to inference on SageMaker. You can learn more about ModelTrainer class https://sagemaker.readthedocs.io/en/stable/api/training/model_trainer.html, ModelBuilder enhancements https://sagemaker.readthedocs.io/en/stable/api/inference/model_builder.html, and get started using https://github.com/aws/amazon-sagemaker-examples/blob/default/%20%20%20%20%20%20build_and_train_models/sm-model_trainer/model_trainer_overview.ipynb and https://github.com/aws/amazon-sagemaker-examples/tree/default/%20%20%20%20%20deploy_and_monitor/sm-model_builder sample notebooks.

Amazon SageMaker enhances search results in Amazon SageMaker Unified Studio with additional context that improves transparency and interpretability. Users can see which metadata fields matched their query and understand why each result appears, increasing clarity and trust in data discovery. The capability introduces inline highlighting for matched terms and an explanation panel that details where and how each match occurred across metadata fields such as name, description, glossary, schema, and other metadata. The enhancement reduces time spent evaluating irrelevant assets by presenting match evidence directly in search results. Users can quickly validate relevance without opening individual assets. This capability is now available in all AWS Regions where Amazon SageMaker is supported. To learn more about Amazon SageMaker, see Amazon SageMaker https://docs.aws.amazon.com/next-generation-sagemaker/latest/userguide/what-is-sagemaker.html. 

Amazon SageMaker Catalog now supports governed classification through Restricted Classification Terms, allowing catalog administrators to control which users and projects can apply sensitive glossary terms to their assets. This new capability is designed to help organizations enforce metadata standards and ensure classification consistency across teams and domains. With this launch, glossary terms can be marked as "restricted", and only authorized users or groups—defined through explicit policies—can use them to classify data assets. For example, a centralized data governance team may define terms like “Seller-MCF” or “PII” that reflect data handling policies. These terms can now be governed so only specific project members (e.g., trusted admin groups) can apply them, which helps support proper control over how sensitive classifications are assigned. This feature is now available in https://docs.aws.amazon.com/sagemaker-unified-studio/latest/adminguide/supported-regions.html where Amazon SageMaker Unified Studio is supported. To get started and learn more about this feature, see SageMaker Unified Studio https://docs.aws.amazon.com/sagemaker-unified-studio/latest/userguide/create-maintain-business-glossary.html.

AWS continues to expand access to the most advanced foundation models with OpenAI open weight models now available in Amazon Bedrock and Amazon SageMaker JumpStart. Accessing these new models from OpenAI on AWS, gpt-oss-120b and gpt-oss-20b, gives you more freedom to innovate and choose the right model for your specific use cases while maintaining complete control over your infrastructure and data.

The first models in the new Llama 4 herd of models—Llama 4 Scout 17B and Llama 4 Maverick 17B—are now available on AWS. You can access Llama 4 models in Amazon SageMaker JumpStart. These advanced multimodal models empower you to build more tailored applications that respond to multiple types of media. Llama 4 offers improved performance at lower cost compared to Llama 3, with expanded language support for global applications. Featuring mixture-of-experts (MoE) architecture, these models deliver efficient multimodal processing for text and image inputs, improved compute efficiency, and enhanced AI safety measures. According to Meta, the smaller Llama 4 Scout 17B model, is the best multimodal model in the world in its class, and is more powerful than Meta’s Llama 3 models. Scout is a general-purpose model with 17 billion active parameters, 16 experts, and 109 billion total parameters that delivers state-of-the-art performance for its class. Scout significantly increases the context length from 128K in Llama 3, to an industry leading 10 million tokens. This opens up a world of possibilities, including multi-document summarization, parsing extensive user activity for personalized tasks, and reasoning over vast code bases. Llama 4 Maverick 17B is a general-purpose model that comes in both quantized (FP8) and non-quantized (BF16) versions, featuring 128 experts, 400 billion total parameters, and a 1 million context length. It excels in image and text understanding across 12 languages, making it suitable for versatile assistant and chat applications. Meta’s Llama 4 models are available in Amazon SageMaker JumpStart in the US East (N. Virginia) AWS Region. To learn more, read the launch blog and technical blog. These models can be accessed in the Amazon SageMaker Studio.

AWS now enables extensive customization of Amazon Nova foundation models through SageMaker AI with techniques including continued pre-training, supervised fine-tuning, direct preference optimization, reinforcement learning from human feedback and model distillation to better address domain-specific requirements across industries.

Amazon SageMaker now offers a unified scheduling experience for visual ETL flows and queries. The next generation of Amazon SageMaker is the center for all your data, analytics, and AI, and includes SageMaker Unified Studio, a single data and AI development environment. Visual ETL in Amazon SageMaker provides a drag-and-drop interface for building ETL flows and authoring flows with Amazon Q. The query editor tool provides a place to write and run queries, view results, and share your work with your team. This new scheduling experience simplifies the scheduling process for Visual ETL and Query editor users. With unified scheduling you can now schedule your workloads with Amazon EventBridge Scheduler from the same visual interface you use to author your query or visual ETL flow. Previously, you needed to create a code-based workflow in order to run a single flow or query on schedule. You can also view, modify or pause/resume these schedules and monitor the runs they invoked. This new feature is now available in all AWS regions where Amazon SageMaker is available. Access the supported region list for the most up-to-date availability information. To learn more, visit our Amazon SageMaker Unified Studio documentation, blog post and Amazon EventBridge Scheduler pricing page.

You can now preview your Amazon S3 Tables directly in the S3 console without having to write a SQL query. You can view the schema and sample rows of your tables stored in S3 Tables to better understand and gather key information about your data quickly, without any setup. You can preview tables in the S3 console in https://docs.aws.amazon.com/AmazonS3/latest/userguide/s3-tables-regions-quotas.html. You only pay for S3 requests to read a portion of your table. See https://aws.amazon.com/s3/pricing/ and the https://docs.aws.amazon.com/AmazonS3/latest/userguide/s3-tables.html for pricing details and to learn more.

Amazon SageMaker Unified Studio announces corporate identity support for interactive Apache Spark sessions through AWS Identity Center’s trusted identity propagation. This new capability enables seamless single sign-on and end-to-end data access traceability for data analytics workflows. Data engineers and scientists can now access data resources in Apache Spark sessions in their JupyterLab environment using their organizational identities, while administrators can implement fine-grained access controls and maintain comprehensive audit trails. For data administrators, this feature simplifies security management using AWS Lake Formation, Amazon S3 Access Grants, and Amazon Redshift Data APIs, enabling centralized access controls across Amazon EMR on EC2, EMR on EKS, EMR Serverless, and AWS Glue. Organizations can define granular permissions based on identity provider credentials for Spark sessions and SageMaker Studio notebook flows, including training and processing jobs. This integration is complemented by comprehensive AWS CloudTrail logging of all user activities—from interactive JupyterLab sessions to https://docs.aws.amazon.com/singlesignon/latest/userguide/user-background-sessions.html - streamlining compliance monitoring and audit requirements. Identity support for Spark sessions in SageMaker Unified Studio is available in the following AWS Regions: US East (N. Virginia), US East (Ohio), US West (Oregon), Canada (Central), South America (São Paulo), Europe (Ireland), Europe (Frankfurt), Europe (London), Europe (Paris), Europe (Stockholm), Asia Pacific (Mumbai), Asia Pacific (Singapore), Asia Pacific (Sydney), Asia Pacific (Seoul), and Asia Pacific (Tokyo). To learn more, visit the https://docs.aws.amazon.com/sagemaker-unified-studio/latest/adminguide/trusted-identity-propagation.html.

AWS announces general availability of Data Lineage in Amazon DataZone and next generation of Amazon SageMaker, a capability that automatically captures lineage from AWS Glue and Amazon Redshift to visualize lineage events from source to consumption. Being OpenLineage compatible, this feature allows data producers to augment the automated lineage with lineage events captured from OpenLineage-enabled systems or through API, to provide a comprehensive data movement view to data consumers. This feature automates lineage capture of schema and transformations of data assets and columns from AWS Glue, Amazon Redshift, and Spark executions in tools to maintain consistency and reduce errors. With in-built automation, domain administrators and data producers can automate capture and storage of lineage events when data is configured for data sharing in the business data catalog. Data consumers can gain confidence in an asset's origin from the comprehensive view of its lineage while data producers can assess the impact of changes to an asset by understanding its consumption. Additionally, the data lineage feature versions lineage with each event, enabling users to visualize lineage at any point in time or compare transformations across an asset's or job's history. This historical lineage provides a deeper understanding of how data has evolved, essential for troubleshooting, auditing, and validating the integrity of data assets. The data lineage feature is generally available in all AWS Regions where Amazon DataZone and next generation of Amazon SageMaker are available. To learn more, visit Amazon DataZone and next generation of Amazon SageMaker.

テキスト検索とセマンティック検索エンジンの台頭によりeコマースや小売業は消費者にとってより簡単に検索できるよ

Unlock efficient large model training with SageMaker HyperPod flexible training plans - find optimal compute resources and complete training within timelines and budgets.