AWS re:Invent 2025 - Architecting the future: Amazon SageMaker as a data and AI platform (ANT351)

Summary of AWS re:Invent 2025 - Architecting the future: Amazon SageMaker as a data and AI platform (ANT351)

The Evolving Data and AI Landscape

• Organizations are heavily investing in AI, but over 75% have seen no meaningful impact on revenue or cost savings
• Key challenges include:
  • Blurring of roles between data engineers, analysts, and scientists
  • Desire to embrace new technologies without ripping out existing investments
  • Need for end-to-end AI governance and trusted data

The SageMaker Unified Studio Approach

• Provides a single platform for data ingestion, processing, analytics, and AI
• Supports different personas and skill sets with a unified development environment
• Integrates with a lakehouse architecture on Amazon S3 using open formats like Apache Iceberg
• Includes a built-in data and AI catalog for governance and security

Empowering Data Producers

• Streamlines data ingestion, ETL, and metadata management
• Automatically generates business-friendly data descriptions using AI
• Provides data quality monitoring and anomaly detection
• Publishes curated data sets to the enterprise catalog

Enabling Data Consumers

• Allows seamless discovery and access to approved data sets and models
• Provides a new native notebook experience with polyglot support
• Integrates pre-trained models and allows customization
• Enables building and deploying AI-powered agents and knowledge bases

Governing the Data and AI Lifecycle

• Integrates a centralized data and AI catalog for secure discovery and access
• Applies guardrails for AI safety, including toxicity, bias, and hallucination detection
• Supports flexible data mesh architectures with centralized control and local autonomy
• Automatically captures lineage and metadata for end-to-end governance

Technical Innovations

• IM-based domains for quick onboarding using existing IAM roles
• New native notebook experience with instant startup and Spark integration
• Fully serverless Apache Airflow for repeatable data workflows

Customer Success: Carrier's Data Modernization Journey

• Addressed data sprawl, governance, and scaling challenges with an Iceberg-based data lakehouse
• Achieved 66% lower implementation costs and 38% improvement in natural language to SQL accuracy
• Key lessons: choose open formats, build governance in from the start, and leverage AI agents responsibly

Key Takeaways

• SageMaker Unified Studio provides a comprehensive platform to address evolving data and AI needs
• Streamlines workflows for data producers, consumers, and governors
• Leverages open formats, AI, and serverless technologies to drive efficiency and governance
• Enables flexible, scalable, and cost-effective data and AI architectures
• Helps organizations unlock the value of data and AI while maintaining control and trust