AI Trainer, Pallon (Computer vision AI model training and image annotation)
Image
ClassificationTrained and evaluated computer vision models by reviewing and annotating image data to detect and classify structural defects in manholes and underground infrastructure. Applied human-in-the-loop quality control by providing precise feedback to improve model accuracy, recall, and classification performance. Supported responsible AI development by flagging inconsistencies, biases, and failure patterns observed during training and evaluation. • Annotated and reviewed image samples for defect detection/classification • Used evaluation rubrics to score outputs across complexity levels and edge cases • Participated in iterative model improvement through performance feedback loops • Assessed model behavior for quality, bias, and recurring failure modes