shaheed Anwar

Conduct frame-by-frame video annotation and timestamp verification for computer vision datasets focused on human-object and robotic physical interactions. Review complex video recordings to tag and categorize fine-grained physical actions, task completion states, and execution errors according to strict project taxonomy guidelines. Systematically identify ambiguities in video edge cases, document feedback to refine annotation guidelines, and help align contributor interpretations across distributed teams. Perform multi-stage quality assurance passes on labeled video datasets, maintaining consistent accuracy ratings (reported as >98%) across complex spatial-temporal evaluation tasks. Manage high-volume workloads independently under strict milestone deadlines while preserving high precision and throughput.

Evaluated machine learning training data for vision-language models and object recognition systems. Applied structured grading rubrics to rate model output accuracy and identify subtle temporal discrepancies between visual actions and generated text descriptions. Collaborated with project leads and quality trainers via asynchronous communication tools to resolve data conflicts and standardize labeling practices across project cycles. Authored concise, structured error reports and summary notes detailing systematic model failures and labeling edge cases.