Precision Image & Video Annotator

Innodata; Washington (District of Columbia); On-site; USD 42,000 - 63,000; Full time

Job summary

Innodata is seeking a diligent image and video annotator to support AI training across multiple programs. You will label data according to a defined ontology, meet quality and throughput targets, and rotate across projects and modalities.

Responsibilities include creating 2D bounding boxes, segmentation masks, keypoints, and 3D/6DoF annotations as needed, plus multi-object tracking and reviewing model-assisted outputs. Attention to detail and strict guideline adherence are essential.

Qualifications

• 1+ years in image or video annotation or equivalent precision work.
• Familiarity with at least one professional annotation platform (e.g., CVAT, Labelbox).
• Understanding of bounding boxes, segmentation masks, keypoints, and tracking.
• Strong visual attention to detail and ability to maintain standards across thousands of frames.
• Ability to follow labeling guidelines exactly and ask precise questions when unclear.

Responsibilities

• Label 2D bounding boxes, masks and keypoints across imagery and video.
• Produce oriented and rotated bounding boxes where needed.
• Create 3D/6DoF annotations on projects requiring them.
• Perform multi-object tracking with consistent identities across frames.
• Review model-assisted outputs and report recurring errors.
• Follow project ontology and escalate ambiguities for guideline updates.

Skills

Annotation experience

Attention to detail

Guideline adherence

Questioning when unclear

Platform familiarity

Education

None required

Tools

CVAT

V7 Darwin

Labelbox

Scale

Job description

Innodata is seeking a diligent image and video annotator to support AI training across multiple programs. You will label data according to a defined ontology, meet quality and throughput targets, and rotate across projects and modalities.

Responsibilities include creating 2D bounding boxes, segmentation masks, keypoints, and 3D/6DoF annotations as needed, plus multi-object tracking and reviewing model-assisted outputs. Attention to detail and strict guideline adherence are essential.