Bedrock Data Automation (Preview): Simplifying Unstructured Data Processing

Bedrock Data Automation (Preview): Simplifying Unstructured Data Processing

• 1.
  © 2025, AmazonWeb Services, Inc. or its affiliates. All rights reserved. © 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved. Amazon Bedrock Data Automation Arnab Sinha Senior Solutions Architect AWS A M A Z O N W E B S E R V I C E S Jean Malha Specialist Solutions Architect – Amazon Bedrock AWS
• 2.
  © 2025, AmazonWeb Services, Inc. or its affiliates. All rights reserved. What’s on deck The multi-modal content challenge What is Amazon Bedrock Data Automation Use Cases and Applications Overview Demo Questions and Resources
• 3.
  © 2025, AmazonWeb Services, Inc. or its affiliates. All rights reserved. Why is multi-modal content hard? Unprecedented scale of content generation Hard to get the desired accuracy ML is promising but requires specialized expertise Lack of standardized tooling Complex postprocessing to adapt and integrate ML output with downstream systems Diverse formats and variations in asset types
• 4.
  © 2025, AmazonWeb Services, Inc. or its affiliates. All rights reserved. Amazon Bedrock Data Automation A generative AI-powered capability to transform multi-modal unstructured content from documents, images, video, and audio into structured data easily, accurately, and at scale.
• 5.
  © 2025, AmazonWeb Services, Inc. or its affiliates. All rights reserved. Key Features Single inference API to handle production scale Built-in Responsible AI with visual grounding, confidence scores, and toxic content detection Simple and intuitive interface to define output schemas and fine-grained business rules Orchestration across state-of-the-art task-specific models and foundation models to generate highly accurate, consistent output Integration with Amazon Bedrock features and Knowledge base
• 6.
  © 2025, AmazonWeb Services, Inc. or its affiliates. All rights reserved. How it works T Y P E S O F O U T P U T T H A T B D A C A N R E T U R N Standard Output Linearized text representation of asset Gen-AI optimized output: reading / viewing order, semantically related output groupings, etc. Controls to optimize output based on downstream systems with simple selection knobs Automatic modality routing based on semantic modality, not just file type Supported Modalities for Standard Output Custom Output Developer supplied schema based on your downstream systems (Blueprint) Supports tasks such as extraction, key and value normalization, transformations, reasoning, splitting and classification Simple NL interface to define business rules and task logic for each field Console-based assistant to create bespoke blueprints in minutes with sample and desired output description Supported Modalities for Custom Output Images Documents Audio Video Images Documents
• 7.
  © 2025, AmazonWeb Services, Inc. or its affiliates. All rights reserved. Key usecases What are the types of applications where multi-modal content processing is needed? Media Asset Analysis & Monetization Intelligent Document Processing Intelligent Speech Analytics Multimodal Intelligent search
• 8.
  © 2025, AmazonWeb Services, Inc. or its affiliates. All rights reserved. Getting Started with BDA is Easy I N P U T S / O U T P U T S Input Asset Amazon Bedrock Data Automation 2 optional Desired Output Instructions Output Response 1 Images Documents Audio Video Standard Output Configuration List of Custom Output Resources (Blueprint) Linearized Text representation of the asset based on configuration Output returned as JSON + additional files if selected in configuration Custom Schema based for each asset based on matched blueprint Output returned as JSON
• 9.
  © 2025, AmazonWeb Services, Inc. or its affiliates. All rights reserved. © 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved. BDA Demo – in console
• 10.
  © 2025, AmazonWeb Services, Inc. or its affiliates. All rights reserved. Next steps H O W T O G E T S T A R T E D 17 How to get started in console Solutions Guidance