Data Annotation Reviewer (QA) — Egocentric Video
workstream • Manila, Philippines
Posted: July 8, 2026
Job Description
Grow With Us
We produce high-quality training data for a leading AI/robotics client building a large-scale egocentric (first-person) video dataset. Our pipeline auto-generates annotations at scale; your job is to review the segments flagged as potentially inaccurate, judge them against the client's guidelines, and correct or reject them so only clean data is delivered.
This is precision quality-assurance work under a strict standard — audited batches must pass at a high accuracy threshold, and low-scoring batches are rejected outright. Your judgment directly affects whether our deliveries get accepted.
Day In The Life
You will review:
- Language annotations — check that written descriptions and action labels accurately, clearly, and consistently describe what's happening in each video, following detailed labeling guidelines.
- Action segmentation & timing — verify that videos are broken into correctly timed, non-overlapping action segments.
- Hand-pose tracking — sanity-check machine-generated hand and motion tracking, flagging errors and implausible results.
- Compliance — flag privacy issues (e.g. faces or personal information) and any content that doesn't meet collection quality or content standards.
Quality standards
- Contribute to a high, continuously audited accuracy target — consistency matters as much as speed, since weak batches can be rejected as a whole.
- Maintain alignment with other reviewers on shared calibration sets.
- Meet the throughput target
Who You Are
- Excellent written and reading English comprehension — you must accurately apply detailed, nuanced guidelines. This is the single most important requirement.
- Sharp attention to detail and the stamina for focused, cognitively demanding review.
- Consistent, well-reasoned judgment on ambiguous, edge-case segments.
- Reliable high-speed internet and a device capable of smooth video playback.
- Comfort in a remote, target-driven environment with regular calibration checks.
Nice to have
- Prior data annotation, labeling, QA, or content-review experience.
- Familiarity with video annotation / labeling tools.
- Exposure to computer vision, robotics, or hand/gesture tracking concepts.
Additional Content
Grow With Us
We produce high-quality training data for a leading AI/robotics client building a large-scale egocentric (first-person) video dataset. Our pipeline auto-generates annotations at scale; your job is to review the segments flagged as potentially inaccurate, judge them against the client's guidelines, and correct or reject them so only clean data is delivered.
This is precision quality-assurance work under a strict standard — audited batches must pass at a high accuracy threshold, and low-scoring batches are rejected outright. Your judgment directly affects whether our deliveries get accepted.
Day In The Life
You will review:
- Language annotations — check that written descriptions and action labels accurately, clearly, and consistently describe what's happening in each video, following detailed labeling guidelines.
- Action segmentation & timing — verify that videos are broken into correctly timed, non-overlapping action segments.
- Hand-pose tracking — sanity-check machine-generated hand and motion tracking, flagging errors and implausible results.
- Compliance — flag privacy issues (e.g. faces or personal information) and any content that doesn't meet collection quality or content standards.
Quality standards
- Contribute to a high, continuously audited accuracy target — consistency matters as much as speed, since weak batches can be rejected as a whole.
- Maintain alignment with other reviewers on shared calibration sets.
- Meet the throughput target
Who You Are
- Excellent written and reading English comprehension — you must accurately apply detailed, nuanced guidelines. This is the single most important requirement.
- Sharp attention to detail and the stamina for focused, cognitively demanding review.
- Consistent, well-reasoned judgment on ambiguous, edge-case segments.
- Reliable high-speed internet and a device capable of smooth video playback.
- Comfort in a remote, target-driven environment with regular calibration checks.
Nice to have
- Prior data annotation, labeling, QA, or content-review experience.
- Familiarity with video annotation / labeling tools.
- Exposure to computer vision, robotics, or hand/gesture tracking concepts.