Case Study: ZEISS achieves fast, accurate document extraction with Diagrid Dapr Agents

ZEISS goes from prototype to production in 2 months with Diagrid

ZEISS, a German optics and optoelectronics manufacturer, faced the challenge of extracting structured optical data from a vast array of unstandardized prescription documents, including handwritten notes and forms in multiple languages. A manual process was not scalable, and incorrect data extraction would result in manufacturing the wrong lens for a patient. To solve this, ZEISS worked with vendor Diagrid to build a production-grade document extraction pipeline using Dapr Agents and Dapr Workflow.

The solution implemented by Diagrid provided the control, reliability, and flexibility ZEISS required, constraining AI within a durable workflow for predictable results. Using Dapr's Conversation API, the system could swap AI models via configuration without code changes. As a result, ZEISS went from prototype to production in just two months with zero labelled training data, achieving recognition accuracy on par with a specialized machine learning system.

ZEISS