ICF helps FDA accelerate the drug labeling review process with AWS machine learning

The FDA's Division of Medication Error Prevention and Analysis reviews 25 to 50 premarket drug labels per reviewer per year through a manual, labor-intensive process, where misinterpreted labeling can lead to medication errors and serious adverse health events. ICF built the Computerized Labeling Assessment Tool (CLAT) using Amazon DynamoDB, API Gateway, Lambda, S3, and Amazon ECS to run open-source ML software (TensorFlow, Tesseract OCR) with sequential transfer learning for object detection, OCR, and computer vision. CLAT enables multiple simultaneous checks across labels and automatically tracks results in a centralized location, improving consistency. The page does not state specific time-savings or error-reduction percentages.

AWS is Amazon's cloud platform, used across pharma for genomics compute, SageMaker/Bedrock model training, and healthcare-and-life-sciences-specific services like Amazon HealthOmics.

This sits in the Build stage of the AI use-case lifecycle: Named examples from the foundation-model labs and platforms pharma teams use to actually construct, train and generate the model or application.