AWS re:Invent 2025 - What's new with Amazon SageMaker in the era of unified data and AI (ANT216)

Summary of AWS re:Invent 2025 - What's new with Amazon SageMaker in the era of unified data and AI (ANT216)

Overview of Next Generation Amazon SageMaker

• AWS introduced the next generation of Amazon SageMaker at re:Invent 2024 to address key challenges faced by customers:
  • Rapidly evolving data and AI landscape requiring manual stitching of complex tool sets
  • Data sprawl and fragmented governance leading to difficulty leveraging data for AI
  • Need to integrate data analytics and AI at scale across teams and workflows
• The next generation of SageMaker aims to provide a unified experience by:
  1. Offering a SageMaker Unified Studio with integrated tools for SQL, Python, visual workflows, and natural language interaction
  2. Providing a SageMaker Catalog for centralized data and AI governance and access control
  3. Building on an open data lakehouse architecture using Apache Iceberg for consistent data access

SageMaker Unified Studio

• Unified Studio removes the pain of separate development experiences by integrating various tools and interfaces:
  • Code editors for Python, SQL query editors, visual drag-and-drop interfaces
  • Natural language interaction through a generative AI assistant
  • Seamless integration with underlying AWS data processing and compute services
• Enables teams to collaborate without context switching, with all data accessible in one place
• Supports a wide range of user preferences and skill levels

SageMaker Catalog for Governance

• SageMaker Catalog provides a single place for data and AI governance with access controls
  • Allows secure discovery and access to approved data models, compute, and AI assets
  • Provides semantic search, data quality monitoring, sensitivity detection, and end-to-end lineage
  • Integrates Amazon Bedrock guardrails for responsible AI, including blocking sensitive information and reducing hallucinations

Open Data Lakehouse Architecture

• SageMaker is built on an open data lakehouse architecture using Apache Iceberg
  • Enables consistent data access across S3, Redshift, and other Iceberg-compatible engines
  • Provides 15+ zero-ETL integrations to bring in data from operational and enterprise applications
  • Allows federated queries to access on-premises and third-party data sources

Key Launches and Enhancements

1. Visual Workflows and Job Monitoring for Data Engineers:

   • Expanded visual ETL interface to automate Glue jobs and other tasks
   • Allows data engineers to create and visualize complex pipelines without Python DAG code

2. Unstructured Data Support in SageMaker Catalog:

   • Brought S3 buckets into the SageMaker Catalog to manage unstructured data assets
   • Enables data producers to enrich unstructured data with metadata for data consumers

3. QuickSight Integration for Data Analysts:

   • Allows data analysts to quickly visualize data sets in SageMaker and publish dashboards
   • Integrates with the SageMaker Catalog for discovery and access control

4. S3 Table Governance with Tag-based Access Control:

   • Expanded tag-based access control to S3 tables, Redshift, and federated data sources
   • Enables data administrators to define permissions based on tags, scaling access control

Recent Launches

1. One-click Onboarding with IAM-based Domains:

   • Allows users to leverage existing IAM roles to set up SageMaker Unified Studio in minutes
   • Automatically configures access to S3, Glue Data Catalog, and other resources based on the IAM role

2. Serverless Notebook Experience:

   • Introduced a new serverless, web-based notebook environment for interactive data exploration and model development
   • Supports both Python and SQL natively, with automatic data transfer between the two

3. SageMaker Data Agent:

   • A purpose-built AI agent that leverages metadata to provide contextual guidance and code generation
   • Helps users explore data, visualize insights, and build end-to-end ML workflows without manual setup

4. Amazon Athena for Apache Spark Engine:

   • A new high-performance query engine that powers the serverless notebook experience
   • Provides scalable processing of data in the SageMaker lakehouse architecture

Customer Feedback and Impact

• Customers like Deloitte have seen faster delivery and experimentation with the developer-focused, unified SageMaker experience
• The new capabilities have enabled users across data engineering, data science, and analytics to collaborate more effectively on data and AI initiatives

Key Takeaways

• AWS has introduced a next-generation Amazon SageMaker platform that unifies data analytics, machine learning, and AI capabilities
• The SageMaker Unified Studio, Catalog, and open lakehouse architecture aim to address challenges around data sprawl, fragmented governance, and the need for integrated data and AI workflows
• Recent launches have focused on improving onboarding, developer experience, and AI-assisted data exploration and model development
• Customers have reported increased productivity and collaboration across data and AI teams leveraging the new SageMaker capabilities