AI Expert – Multimodal Data Labeling & Evaluation
This role focused on advanced data labeling and AI evaluation for multimodal content. The tasks involved analyzing and verifying relationships between images, text, and video content to improve AI accuracy, auditing datasets for correctness, temporal consistency checks, and bias detection across different data types. Additional responsibilities included prompt engineering and the creation of training data for scene description and classification purposes. • Evaluated image, video, and text datasets to ensure model understanding and alignment. • Performed bias and safety audits on multimodal datasets, detecting context-dependent risk factors. • Designed and tested complex prompts to facilitate accurate AI scene or content understanding. • Categorized short texts and performed semantic tagging on images and documents.