Understand this story
On November 14, 2026, AWS Machine Learning published an article stating that Amazon SageMaker HyperPod has added a Data Capture capability, which can record the inputs and outputs of online inference to Amazon S3 at three independent levels—the endpoint, the load balancer, and the model Pod—with zero impact on model availability. Using Qwen2.5-3B-Instruct home appliance customer service routing as an example, the article demonstrates collecting production traffic, using Athena queries to discover blind spots for new products, labeling with SQL policies and fine-tuning, and then comparing accuracy before and after fine-tuning. There are currently no subsequent developments.
AI-generated from public reports
Latest development · Report
Building a Model Improvement Flywheel with Amazon SageMaker HyperPod Data CaptureDevelopments
1 confirmed developments
- Source:AWS Machine Learning
Reports and discussion timeline
Browse reports and discussion by time. Their evidence levels remain distinct.
Nov 14
AWS Machine LearningFirst-party
Building a Model Improvement Flywheel with Amazon SageMaker HyperPod Data Capture
Amazon SageMaker HyperPod introduces the new Data Capture capability, which can record the inputs and outputs of online inference to Amazon S3 at three independent levels—the endpoint, the load balancer, and the model Pod—with zero impact on model availability. Using Qwen2.5-3B-Instruct home appliance customer service routing as an example, the article demonstrates collecting production traffic, using Athena queries to discover blind spots for new products, labeling with SQL policies and fine-tuning, and then comparing accuracy before and after fine-tuning.