All news with #amazon sagemaker ai tag

🚀 Amazon SageMaker Studio notebooks now support EC2 G7 instances powered by NVIDIA RTX PRO 4500 Blackwell Server Edition GPUs. G7 offers up to 4.6x AI inference performance versus G6, with up to 8 GPUs and 700 Gbps EFA-enabled bandwidth to accelerate inference, graphics, and analytics workloads. These instances are available in AWS US East (N. Virginia and Ohio) and US West (Oregon). Refer to developer guides for JupyterLab and CodeEditor setup and the pricing page for cost details.

🔥 Amazon SageMaker now supports serverless full fine-tuning for more than 25 open-source models, including families such as gpt-oss, Gemma, Llama, Nemotron, and Qwen. In addition to parameter-efficient methods like LoRA, SageMaker customers can update all model parameters to achieve deeper domain adaptation when needed. The service manages infrastructure provisioning and training orchestration, and billing is usage-based.

🔗 Amazon SageMaker Unified Studio now supports Teradata Vantage as a data source, enabling users to query and analyze enterprise data warehouse assets alongside other data in a single governed environment. This integration lets you combine Teradata data with sources such as Amazon Redshift, Amazon S3, and relational databases to correlate analytical and operational workloads. Data engineers and analysts can query Teradata data in the data explorer, use the query editor and notebooks, or include it in visual ETL jobs without leaving the studio.

🔧 Amazon SageMaker Unified Studio now provides full Git version control directly within core tools like Query Editor, Visual ETL, Workflows, and Notebooks. The updated Repositories experience replaces automatic sync with flexible, file-level control and lets teams choose which files to track in Git on GitHub, GitLab, or Bitbucket. Repositories are decoupled from project creation, support multiple repos and branches concurrently, and preserve CLI access via JupyterLab or Code Editor terminals.

🔎 Amazon SageMaker Unified Studio now supports Amazon OpenSearch as a data source, allowing you to query and analyze search and log analytics data alongside other assets. You can combine OpenSearch data with sources like Amazon Redshift, Amazon S3, and relational databases within a single governed environment. The integration enables correlating operational search data with analytical datasets for cross-source insights and streamlined workflows.

📈 Amazon Managed Service for Prometheus now supports up to 1.5 billion active metric time series and up to 200,000 recording and alerting rules per workspace, with customers able to create many workspaces per account. Amazon Managed Service for Prometheus is a fully managed, Prometheus-compatible monitoring service that automatically scales ingestion and storage for high-cardinality workloads across containerized, serverless, and hybrid environments. It integrates with AWS security services to provide secure access to monitoring data and lets customers request higher workspace limits via AWS Support Center or AWS Service Quotas.

🛠️ Amazon RDS for SQL Server now supports Microsoft SQL Server 2025 across Enterprise, Standard, and Developer editions. SQL Server 2025 integrates AI into the database engine, enabling T-SQL to call external REST endpoints and work with AWS services like Amazon Bedrock, Amazon SageMaker, Amazon S3, and AWS Lambda. Standard Edition capacity and features have been expanded, and a new free Developer edition is available for testing.

🔧 With Amazon SageMaker Unified Studio, teams can now create, share, and reuse custom visual transforms within visual ETL flows. This feature enables data engineers to encapsulate business-specific transformation logic—such as phone number standardization, PII masking, or data-quality checks—into reusable components that non-coders can apply across ETL jobs. The capability is available in all AWS Regions where Unified Studio is offered.

🧭 Amazon SageMaker HyperPod now supports partition-level network topology configuration for Slurm clusters, allowing a single cluster to run different topologies (tree or block) per partition. HyperPod auto-detects topology from instance types—UltraServer types map to block and hierarchical-interconnect types map to tree—while partitions without topology data remain schedulable. The feature is enabled by default for Slurm 25.11+ and is available in all Regions that support HyperPod.

🛡️ Amazon GuardDuty introduces AI Protection to extend threat detection to AWS AI services such as Amazon Bedrock and Amazon SageMaker. The feature continuously monitors AI workloads for threats like anomalous invocations, cost harvesting attacks, and prompt injection, using CloudTrail management and data events to surface suspicious activity. Findings integrate with AWS Security Hub for centralized triage and can be enabled per account or centrally via AWS Organizations, with a 30-day trial available for GuardDuty customers.

🛠️ Amazon Managed Service for Apache Flink now includes AI Agent Skills that provide expert, up-to-date guidance for building and operating Flink applications. The skills cover common tasks like creating applications, troubleshooting, scaling, monitoring, networking configuration, and cost optimization. Customers can use these skills with existing AI coding agents such as Kiro, Claude Code, or Cursor by configuring the Agent Toolkit for AWS via the AWS CLI. The feature aims to speed development, simplify upgrades to versions like Flink 2.2, and help maintain healthy, performant streaming applications.

🔒 Amazon Web Services has added OpenAI’s privacy-filter to Amazon SageMaker JumpStart, offering a bidirectional token-classification model for PII detection and masking. The model is designed for fast, context-aware, tunable, high-throughput data sanitization workflows that can run on-premises. It detects PII spans like account numbers, addresses, emails, names, phones, URLs, dates, and secrets, labeling inputs in a single forward pass. Customers can deploy the model via the SageMaker Studio Models section or the SageMaker Python SDK.

🎙️ AWS added Voxtral-Mini-4B-Realtime-2602 to Amazon SageMaker JumpStart, a multilingual, low-latency speech-transcription model from Mistral AI. The model offers natively streaming architecture for real-time transcription across 13 languages and configurable delay/accuracy trade-offs. Customers can deploy it via the SageMaker Studio Models section or the SageMaker Python SDK for rapid integration into speech applications.

🔍 AWS added Qwen3-VL-Embedding-2B and Qwen3-Reranker-4B to Amazon SageMaker JumpStart, enabling modular search and retrieval pipelines. The embedding model creates multimodal vectors from text, images, screenshots, and video across 30+ languages, while the reranker scores query–document pairs across 100+ languages. Customers can deploy these models via SageMaker Studio or the SageMaker Python SDK with a few clicks.

📰 Amazon Web Services has added gemma-4-E2B-it, a multimodal instruction-tuned foundation model from Google DeepMind, to Amazon SageMaker JumpStart. The model supports text, image, audio inputs and text outputs with a built-in stepwise reasoning mode and broad capabilities including image and video understanding, OCR, function calling, and multilingual code assistance. Customers can deploy it from the SageMaker Studio Models section or via the SageMaker Python SDK for rapid integration into AWS AI workflows.

🔒 Amazon SageMaker HyperPod now supports custom AMIs for Slurm-orchestrated clusters, enabling deployment of pre-configured, security-hardened environments tailored to organizational needs. This reduces reliance on complex lifecycle scripts and improves startup times and consistency across nodes. Custom AMIs must be built from HyperPod's public base AMIs for compatibility and can be specified via the CreateCluster, UpdateCluster, or UpdateClusterSoftware APIs. The feature is available in all Regions where HyperPod is supported.

🛠️ Amazon EMR on EKS now integrates an Apache Spark troubleshooting agent that provides automated root cause analysis and PySpark recommendations through natural language, simplifying diagnosis of job failures. The agent inspects Spark History Server data, executor logs, and cluster configs to detect issues like memory errors, data skew, resource contention, and connectivity problems. Accessible via a "Troubleshoot with AI" option in the EMR on EKS console and via MCP with compatible AI coding agents, the feature is read-only, IAM-authenticated, logged in CloudTrail, and available in Regions with SageMaker Unified Studio.

🚀 Amazon SageMaker HyperPod now supports AMI-based configuration for Slurm clusters that use continuous provisioning, enabling nodes to be provisioned with required software and settings without managing lifecycle scripts in S3. AMI-based configuration installs components like Docker, Enroot, and Pyxis and applies Slurm accounting, SSH key management, and log rotation as nodes are added. To enable it, omit the LifeCycleConfig block via the API or select "None" under Lifecycle scripts in Custom setup in the console; an optional extension script can be provided via OnInitComplete and SourceS3Uri. This feature is available in all AWS Regions where SageMaker HyperPod is offered, and custom lifecycle scripts remain supported for advanced use cases.

🛠️ Amazon SageMaker Unified Studio now supports custom asset types for IAM-based domains, allowing administrators to catalog diverse asset formats such as medical images, PowerBI dashboards, or PDF reports. Administrators define a type with name, description, and optional metadata forms; assets created from those types can include glossary terms and README documentation. Published assets are discoverable and subscribable through the same governed workflow used for other catalog items and are available in all Regions where Unified Studio is offered.

🔍 Amazon SageMaker HyperPod now supports deep health checks for Slurm clusters created with continuous provisioning, enabling proactive verification of GPU accelerator health on running instances. Continuous provisioning allows asynchronous scaling of instance groups without all-or-nothing failures, and deep health checks validate hardware as nodes come online. Results and progress are visible via the SageMaker console and APIs, and failing instances are isolated and recovered automatically.