Artificial Intelligence Implementor

About the Role

We build AI-powered turnaround monitoring for airports. Our cameras are installed across parking stands to detect and classify ground turnaround events chocks, GPU, doors, fuelling, boarding displayed live on our mobile app and OCC dashboard used by airline operations teams.

The Annotation Analyst owns the data pipeline that powers our models. You will plan annotation tasks per project, manage the annotator team, ensure quality, and coordinate with ML engineers to keep training datasets on schedule.

Key Responsibilities

• Plan annotation requirements for each AI model volume, event types, camera views, and format
• Assign tasks to annotators and track daily/weekly progress across all active projects
• Review annotation quality; give structured feedback and enforce standards
• Manage annotator workload, onboarding, and performance
• Coordinate with ML engineers to ensure datasets meet training requirements
• Maintain trackers showing annotation progress vs model training milestones
• Own annotation tooling setup and project configuration (CVAT / Label Studio or similar)
• Flag risks early when annotation timelines are at risk of slipping

Requirements

Must Have

• 35 years in an annotation, AI data operations, or ML data management role
• Hands-on experience with computer vision annotation bounding boxes, activity classification, object tracking
• Experience managing or supervising an annotation team
• Strong planning and tracking skills across multiple concurrent projects
• Proficient with at least one annotation tool CVAT, Label Studio, Roboflow, or similar

Good to Have

• Video annotation experience frame selection, event-based labelling, temporal tagging
• Familiarity with aviation operations or airport ground handling
• Exposure to ML model training workflows

What Success Looks Like

• Annotation plans exist for every active AI project before model training begins
• Annotator team runs to a clear, measurable weekly cadence
• ML engineers are unblocked no delays caused by missing or poor-quality data
• Quality issues are caught before data reaches model training