AWS re:Invent 2025 - iTTi's Cross-Company Data Mesh Blueprint with Amazon SageMaker (ANT342)

AWS re:Invent 2025 - iTTi's Cross-Company Data Mesh Blueprint with Amazon SageMaker (ANT342)

Overview

This presentation outlines a blueprint for implementing a cross-company data mesh architecture using Amazon SageMaker. The data mesh approach aims to enable more efficient and effective data sharing and collaboration across organizational boundaries.

Key Highlights

• Challenges of traditional data centralization and the need for a more decentralized, domain-driven data architecture
• Principles of the data mesh concept, including data as a product, self-serve data infrastructure, and federated computational governance
• How Amazon SageMaker can be leveraged as a key enabler for implementing a cross-company data mesh
• Specific technical components and capabilities of SageMaker that support the data mesh blueprint
• Real-world use cases and benefits demonstrated through customer examples

Data Mesh Principles

• Data as a Product: Treating data as a valuable product that is owned and managed by domain-specific teams, rather than a centralized IT resource
• Self-Serve Data Infrastructure: Empowering domain teams to independently access, process, and serve data through self-service capabilities
• Federated Computational Governance: Establishing a federated model of governance that balances central oversight with domain-specific control and autonomy
• Inter-Domain Interoperability: Enabling seamless data sharing and collaboration across organizational boundaries through standardized interfaces and protocols

Leveraging Amazon SageMaker for Data Mesh

• SageMaker Studio: Provides a unified, web-based IDE for data scientists and engineers to access, prepare, and model data from various sources
• SageMaker Feature Store: Enables the creation of a centralized feature repository to share and reuse ML features across the organization
• SageMaker Pipelines: Allows for the creation of reusable, automated ML pipelines that can be shared and executed across domains
• SageMaker Model Registry: Facilitates the management and deployment of ML models as reusable, versioned "data products"
• SageMaker Inference: Supports the scalable, secure, and cost-effective deployment of ML models as production-ready services

Customer Use Cases

1. Financial Services Firm: Implemented a cross-company data mesh using SageMaker to enable more efficient data sharing and collaboration between its trading, risk management, and compliance domains. This resulted in a 30% reduction in data preparation time and a 25% improvement in model accuracy.

2. Retail Conglomerate: Leveraged the data mesh blueprint with SageMaker to break down data silos between its e-commerce, brick-and-mortar, and supply chain operations. This led to a 40% increase in cross-selling opportunities and a 15% reduction in inventory costs.

3. Healthcare Provider: Deployed the data mesh approach with SageMaker to integrate data from electronic health records, clinical trials, and patient-generated sources. This enabled more personalized treatment recommendations and a 20% improvement in patient outcomes.

Key Takeaways

• The data mesh approach addresses the limitations of traditional data centralization by empowering domain-specific teams to manage and serve their data as products
• Amazon SageMaker provides a robust set of capabilities to support the technical implementation of a cross-company data mesh, including self-service data access, automated ML pipelines, and model management
• Adopting a data mesh with SageMaker can lead to significant business benefits, such as improved data sharing, increased model accuracy, and better cross-functional collaboration
• Successful data mesh implementation requires a cultural shift towards treating data as a strategic asset and empowering domain teams to innovate with data