Opportunity brief
About the Company
Praxis Robotics is an AI company that gathers, processes, and monetizes task data from real-world work environments. We record manual tasks performed across various industries, strip out all personally identifiable information, and rigorously annotate the footage to build high-fidelity datasets. These annotated datasets are then sold directly to frontier labs to train physical AI, robotics, and advanced perception models.
Responsibilities
- Evaluate, annotate, and perform quality control on video footage and potentially sensor outputs, ensuring data meets certain criteria and pre-generated labels are correct.
- Review and correct multi-pass automated quality checks and machine-generated outputs—ensuring accuracy across broad task goals, subtask breakdowns, task completion status, and baseline metadata (environment, task type, success/failure tags).
- Verify visual clips against strict baseline requirements (e.g., active work vs. idle time, hands continuously in frame, camera lighting, sharpness/blur, and general visibility).
- Serve as the primary human validation layer for complex or high-fidelity datasets, taking over automated passes to fix edge cases and maintain strict quality thresholds.
- Document common failure modes, log edge cases, and provide feedback to refine automated pipeline checks and annotation guidelines.
Requirements
- Strong attention to detail, visual discipline, and commitment to producing accurate work.
- Reliable workstation (laptop/PC) with high-speed, stable internet access.
- High proficiency in written English communication.
- University student, recent graduate, or equivalent background (STEM, Computer Science, or analytical fields preferred).
- Prior Data Annotation Experience: Experience labeling or reviewing images, video, or multimodal data.
- Familiarity with AI review processes, quality assurance frameworks, and executing complex scoring rubrics.
- Proactive problem-solving skills and self-motivated work style in a remote setting.