#49426 Video Annotation & Data Labeling Project at Mindy

We’re looking for detail-oriented specialists to join a long-term AI data annotation project focused on high-level video captioning, behavioral segmentation, and AI model verification.

Successful contributors will begin with video annotation and segmentation tasks and gain priority access to advanced, high-paid projects.

What you will do:

• High-Level Captioning: Write 1–2 sentence summaries of video action sequences in simple English. You must strictly follow objective guidelines (no mentions of humans, body parts, or assumptions).

• Video Segmentation: Break down video timelines into discrete, action-based segments (minimum 2–3 per task) ensuring zero time gaps or overlaps.

• Action Labeling: Describe visible actions using simple present tense (e.g., "Picks up the object"). Labels must be strictly based on what is visually verifiable.

• Model Interaction & QA: Review and edit automatically generated captions from automated AI models to ensure 100% accuracy and consistency.

Qualifications needed: 

• High Accuracy
• Deliver high-quality annotations with 85–90%+ benchmark accuracy
• Follow annotation guidelines consistently
• Describe only what is visible—without assumptions or interpretations
• Use clear, consistent language
• Productivity
• Complete the annotation workflow for a 1-minute video in approximately 4 minutes
• Efficiently use keyboard shortcuts
• Run background screen-recording software during tasks
• Professional Work Style
• Comfortable working in a structured, quality-driven environment
• Able to stay focused throughout task execution
• Follow project processes and productivity requirements 

Certification & Onboarding Process

To ensure all specialists meet the required standards, candidates must pass a certification process (approx. 3 hours total):

1. Preparation: Review guidelines and training videos.

2. Practice: Master the shortcuts and rules on available training tasks.

3. Certification Test: Successfully complete 5 certification tasks by delivering high-quality results within the 4-minute time limit per video.