AWS re:Invent 2025 - Enabling AI innovation with Amazon SageMaker Unified Studio (ANT352)

Enabling AI Innovation with Amazon SageMaker Unified Studio

Building a Strong Data Foundation

• Importance of establishing a solid data foundation for successful AI and analytics initiatives
• Key challenges organizations face: lack of trust in data, insufficient data literacy, and absence of a data-driven culture
• Two core aspects of building a data foundation:
  1. Improving data literacy within the organization
  2. Cultivating a robust data culture

Enhancing Data Literacy

• Seven key data actions to improve data literacy:
  1. Building a comprehensive data inventory
  2. Annotating data with business context
  3. Detecting and categorizing PII data
  4. Assessing data quality
  5. Establishing data lineage
  6. Publishing new data assets
  7. Choosing a customized data literacy improvement journey

Fostering a Data-Driven Culture

• Parallels between building a social platform and a data culture
  • Producers (content creators), consumers (users), and a central team (platform providers)
• Importance of making data culture a part of the organization's DNA and sustaining it through executive changes

Amazon SageMaker: Enabling Data and AI Governance

• Overview of the three-layer architecture of Amazon SageMaker:
  1. Lakehouse architecture
  2. Data and AI governance capabilities
  3. Unified data experience
• Detailed look at the data and AI governance features within Amazon SageMaker:
  • PII detection, data quality checks, business data catalog, and more
• Flexibility to start the journey with any layer of the SageMaker platform

Commonwealth Bank of Australia's Data Transformation Journey

Decentralized, Federated Data and AI Strategy

• Three core pillars of CBA's data and AI strategy:
  1. Empowering people through decentralization and embedding data teams
  2. Implementing robust governance and controls
  3. Leveraging technology, particularly the partnership with AWS

Migrating to the AWS Cloud

• Challenges of moving from an on-premises Hadoop-based data lake to the cloud
• Lift-and-shift approach to migrate 10+ PB of data to Amazon S3
• Enabling access to diverse cloud-native analytics and ML engines

Building a Federated Data Mesh Platform

• Transitioning from a monolithic to a composable, decentralized platform
• Implementing an abstraction layer to unify access and governance across multiple AWS accounts
• Integrating Amazon SageMaker Unified Studio for a seamless user experience

Key Outcomes and Learnings

• Empowering teams with self-service data access and a unified dashboard
• Streamlining workflows and accelerating experimentation
• Recognizing the importance of MVPs, steel threads, and iteration
• Embracing the learning curve and the unknown during transformation

Demonstration: Enabling Governed AI Innovation

• Multi-account environment setup with shared services, data producer, and data consumer accounts
• Walkthrough of the end-to-end user experience in the SageMaker Unified Studio:
  • Data discovery, access request, and approval workflow
  • Automated data quality checks and metadata enrichment using generative AI
  • Building and training a fraud detection model leveraging the federated data platform

Key Takeaways

1. Building a strong data foundation and data-driven culture is crucial for successful AI and analytics initiatives.
2. Amazon SageMaker provides a comprehensive platform for data and AI governance, enabling organizations to start their journey at any layer.
3. Commonwealth Bank of Australia's data transformation journey demonstrates the power of a federated, decentralized data platform built on AWS.
4. Embracing MVPs, steel threads, and iteration is key to navigating the challenges of data and AI transformation.
5. Generative AI can significantly streamline data management tasks, such as metadata enrichment and data quality checks.