Data Annotator / AI Trainer & Atlas Support (Remote) (Atlas Capture)
Served as an AI Trainer and Data Annotator supporting human action labeling for machine learning training. Performed atomic action decomposition using Atomic Action Labels (AAL) with frame-level and clip-level temporal segmentation. Conducted review, correction, and quality assurance to maintain dataset integrity and consistency across large production datasets. • Labeled and decomposed complex activities into atomic units using AAL. • Completed 3300+ footages totaling 4000+ minutes with 94% production accuracy. • Marked action start and end points at frame and clip levels for precise temporal alignment. • Reviewed flagged/ambiguous edge-case clips, corrected label drift, and provided secondary QA feedback.