AWS adds Claude tool use to Bedrock for custom entity extraction from documents
Amazon Web Services has integrated Claude's tool use (function calling) capability into Bedrock, enabling serverless document processing for custom entity recognition. The solution uses Claude 3.5 Sonnet to extract structured data like names, dates, and addresses from driver's licenses and other documents without traditional model training.
AWS Adds Claude Tool Use to Bedrock for Custom Entity Extraction
Amazon Web Services has published a production-ready implementation guide for using Anthropic's Claude model with tool use (function calling) capabilities in Amazon Bedrock to extract structured data from unstructured documents.
What's New
The solution demonstrates how to build a serverless document processing pipeline that extracts custom fields from documents like driver's licenses without requiring model training or fine-tuning. Claude evaluates prompts and automatically determines whether to invoke predefined tools to complete extraction tasks.
Technical Implementation
The architecture combines AWS services into an automated workflow:
- Document Upload: Users upload documents to Amazon S3
- Event Triggering: S3 PUT event automatically triggers an AWS Lambda function
- Model Invocation: Lambda sends the document to Claude via Amazon Bedrock's invoke-model API
- Entity Extraction: Claude processes the image and extracts fields using defined tool schemas
- Monitoring: Results are logged in Amazon CloudWatch
The implementation uses Claude 3.5 Sonnet (global.anthropic.claude-sonnet-4-5-20250929-v1:0) with the tool_choice: auto parameter, allowing Claude to decide whether tool invocation is necessary. For deterministic behavior in production, developers can hardcode specific tool selection.
Key Technical Details
Tool Definition: Developers define tools with JSON schemas specifying field names, types, and requirements. The example extracts driver's license fields including first name, last name, issue date, license number, and address components.
Lambda Configuration: The guide recommends setting Lambda timeout to 30 seconds minimum (1-2 minutes for high-resolution images) to accommodate Claude's processing latency. Default 3-second timeouts will cause failures.
Image Processing: Documents are base64-encoded and sent with text prompts to Claude. The model returns structured JSON matching the defined tool schema.
IAM Permissions: Required permissions include bedrock:InvokeModel for Claude access and s3:GetObject for document retrieval.
Capabilities
The tool use approach supports:
- Dynamic field extraction without retraining
- Multiple document types (driver's licenses, insurance cards, ID badges, business forms)
- Scalable, serverless processing
- Real-time extraction with no batch processing required
- Flexible schema definition based on use case
What This Means
This integration reduces barrier to entry for document automation. Previously, extracting data from images required either rule-based OCR systems, fine-tuned models, or manual setup. Claude's tool use with vision capabilities removes the training requirement—developers can define extraction schemas in minutes and deploy production pipelines immediately. The serverless architecture eliminates infrastructure management while AWS Lambda's automatic scaling handles variable document volumes. Organizations processing hundreds or thousands of documents can now automate extraction without building custom ML pipelines.
Related Articles
AWS Overhauls Bedrock AgentCore Runtime, Cuts Cold Starts to Flat 2 Seconds Regardless of Image Size
Amazon has released an updated Bedrock AgentCore runtime that holds cold start latency at roughly 2 seconds regardless of container image size, versus up to 30 seconds on the previous version. The update also changes memory billing to track real-time usage instead of peak allocation.
xAI's Grok 4.6 Launches on Amazon Bedrock With 500K Context and Cross-Region Inference
xAI's Grok 4.6 is now available on Amazon Bedrock via both bedrock-mantle and bedrock-runtime endpoints, adding Converse API support, cross-Region inference profiles, and Bedrock Guardrails. The model offers a 500K token context window and four reasoning effort levels, with input pricing starting at $2.00 per million tokens on the global inference profile.
AWS Ships 13 SageMaker Inference Features in 2026, Cutting Startup Latency 51% and GPT-OSS-20B Throughput 2x
Amazon rolled out 13 new SageMaker AI inference capabilities in 2026 across managed endpoints and HyperPod Inference, spanning automated benchmarking, instance-pool fallback, OpenAI-compatible APIs, and container caching. AWS claims container caching cut endpoint startup latency by 51% and an inference-recommendation feature doubled GPT-OSS-20B throughput at equal latency.
Wood Mackenzie Builds Shared Agentic Platform APEX on Amazon Bedrock AgentCore
Wood Mackenzie built APEX (Agentic Platform for Energy eXperience) on Amazon Bedrock AgentCore to give three separate applications a shared runtime for identity, guardrails, memory, and scaling instead of each rebuilding the same infrastructure. The company says 88% of its internal AI proofs-of-concept never reach wide deployment, a gap it attributes to architecture rather than model quality.
Comments
Loading...