Video Annotation Reviewer

You'll audit video annotation work submitted by other labelers — verifying that temporal labels, content tags, and moderation flags have been applied correctly and consistently. A typical task gives you a short video clip alongside the annotation a labeler submitted, and your job is to confirm accuracy, correct errors, and escalate edge cases that fall outside the label taxonomy. You're the last human checkpoint before this data goes into a training pipeline, so precision matters more than speed.

The annotation types you'll review include action recognition labels (start/end timestamps for specific activities), content moderation flags (violence, adult content, misinformation indicators), and English-language transcription segments matched to audio. You'll work through a structured interface that shows the original clip, the submitted labels, and fields for your verdict and notes. Disagreements with the original labeler require a written rationale. You are not re-labeling from scratch — you're auditing existing work, which means you need to be both accurate and consistent with the project's labeling guidelines.

You need at least 2 years of video annotation or content moderation experience, with strong English comprehension for transcript-verification tasks. Familiarity with timestamp-based annotation tools (e.g., CVAT, Label Studio, or proprietary tooling) is expected. Moderation experience with platforms that handle sensitive content is a plus, as some batches include flagged material requiring careful classification.