Case Study: GoCheck Kids boosts vision screening accuracy with Provectus ML infrastructure

GoCheck Kids enables faster ML experimentation, increases the recall of ML models by 3x

GoCheck Kids, a clinically validated pediatric vision screening app, wanted to improve its image classification and help users avoid “not gradable” results caused by issues like a child not looking at the camera. To do this, the company needed a robust, scalable machine learning infrastructure that could handle experiments on more than 1 million images and support faster, more cost-efficient model training and refinement.

Provectus partnered with GoCheck Kids to build a secure, auditable, and reproducible ML training environment on AWS, using Kubeflow and Amazon SageMaker with experiment tracking, model versioning, and active learning pipelines. The result was a 3X increase in recall while preserving precision, plus the ability for three ML engineers to run 100+ large-scale experiments in three weeks, with 95% of their time now focused on experimentation.

GoCheck Kids