AI Labelling
The project involved designing labeling guidelines, configuring annotation tasks on the Toloka platform, managing contributor workflows, and implementing multi-level quality control mechanisms to ensure dataset accuracy and consistency. Tasks included image classification, text annotation, and validation workflows depending on model requirements. I structured task instructions to reduce ambiguity, optimized task pricing and batching to improve turnaround time, and monitored annotator performance using built-in Toloka analytics. Quality assurance processes included golden tasks, majority voting, manual review layers, and performance-based contributor filtering. The labeled data was exported, cleaned, and formatted for integration into downstream machine learning pipelines, enabling improved model performance and reliable training data standards.