Project Cursa - Robot Manipulation Video Annotator

Overview

Welo Data is looking for detail-oriented annotators to label robot manipulation videos for AI training. The work involves reviewing short videos captured from three synchronized camera angles and writing precise, structured natural-language descriptions of the robot’s actions. This is remote, project-based freelance work.

What You'll Do

• Watch short robot manipulation videos from an overhead camera and two wrist-mounted camera views.
• Split each video into time segments and write clear descriptions for each segment.
• Apply labels at three levels of detail when relevant: atomic motion, skill/subtask, and task/goal.
• Ensure the full video is covered with labels at two or more levels, including idle or pause moments.
• Describe exactly what happens, including failed grasps, dropped objects, slips, and other non-ideal outcomes.
• Use all three camera angles together to confirm the scene, object identity, contact points, and grasp details.
• Follow a detailed style guide for vocabulary, spatial relations, object parts, and movement descriptions.
• Participate in periodic calibration sessions to keep labeling aligned with the team and reference examples.

Requirements

• Strong written English with the ability to produce many short, precise sentences without repeating the same phrasing.
• Sharp attention to detail, especially for small distinctions in robot motion and object interaction.
• Comfort following a detailed structured style guide consistently, including in ambiguous cases.
• Basic ability to describe spatial and mechanical relationships, such as left/right, above/below, and object parts like handles or edges.
• Reliable, self-directed work habits for mostly independent, heads-down work with periodic check-ins.

Nice to Have

• Prior experience with video annotation, data labeling, transcription, or QA work.
• Familiarity with robotics terminology such as grippers, end-effectors, and manipulation.
• Experience with annotation tools such as Label Studio.

Note: The source also includes a generic line about applying expertise in the legal field, but the role itself is for robot manipulation video annotation, so that line appears inconsistent with the rest of the posting.