Project Cursa - Robot Manipulation Video Annotation QC (V2)

The work

Review robot manipulation video annotations to maintain accurate, consistent training data for robotics AI. Compare labeled segments with a detailed style guide and reference examples, identify defects, and provide constructive feedback. The role requires judgment about nuanced descriptions, not just typo checking.

Responsibilities

• Review completed labels at all three detail levels: atomic motion, skill/subtask, and task/goal.
• Cross-check descriptions against the overhead camera and both wrist-mounted cameras to verify actions, object identity, contact, and grasp details.
• Confirm full video coverage with no labeling gaps, including idle periods and pauses.
• Verify that descriptions reflect actual outcomes, including failures, drops, and slipped grips. Flag inaccurate or overly favorable accounts of what occurred.
• Check style-guide compliance: action vocabulary, spatial and directional accuracy, consistent object and action terminology, and varied language rather than repetitive phrasing.
• Identify recurring errors or areas of annotator confusion and report them to the team lead for calibration.
• Give specific, actionable feedback on rejected or revised annotations, citing the style guide and reference examples.
• Attend periodic calibration with the team and client as guidelines evolve or edge cases are resolved.
• Maintain a QC log or tracking sheet when required, recording review outcomes, common error types, and annotator performance trends.

Required skills

• Excellent written English and strong reading comprehension for precise, nuanced language evaluation.
• Exceptional attention to subtle mismatches, such as the wrong grasp type, an incorrect object part, or an overlooked failure event.
• Strong spatial and mechanical description skills, including left/right, above/below, and object parts such as handles, lids, and edges.
• Consistent use of a detailed, structured style guide and sound judgment in ambiguous or edge-case situations.
• Comfort giving clear, specific, constructive written feedback rather than only pass/fail marks.
• Reliable, self-directed work habits and independent delivery within review-turnaround expectations.

Preferred background

• QC, editing, proofreading, or annotation-review experience.
• Video annotation, data labeling, or transcription QA experience.
• Familiarity with robotics terms such as grippers, end-effectors, and manipulation.
• Annotation-tool experience, including Label Studio review and approval workflows.

Arrangement and support

This is fully remote, project-based work with flexibility over when and how much you contribute. Optional AI and Large Language Model workshops, responsive guidance, and a global contributor community are available. No coding is required. Registration starts with a few introductory questions for the contributor database.