Data Annotator

Innodata Inc.•Washington, DC

1d•$42,000 - $62,500•Remote

About The Position

Innodata is a global data engineering company focused on enabling the responsible advancement of artificial intelligence. We provide data, evaluation frameworks, and human expertise to build trustworthy AI systems at scale. Our solutions, platforms, and services cater to Generative AI/AI builders and adopters, leveraging our 36+ year legacy of delivering high-quality data and outstanding customer outcomes. This role involves producing high-fidelity annotations on imagery and video to train and evaluate machine learning models. Annotators will adhere to a defined ontology and labeling specification, meet project quality and throughput targets, and operate within designated annotation environments. Assignments may rotate across projects, modalities, and customers based on program needs.

Requirements

1+ years of image or video annotation experience, or equivalent precision work in a quality-managed production environment.
Working familiarity with at least one professional annotation platform (for example CVAT, V7 Darwin, Labelbox, Scale, or comparable).
Practical understanding of bounding boxes, segmentation masks, keypoints, and object tracking — and of what makes each one correct rather than merely present.
Strong visual attention to detail and the discipline to hold a standard across tens of thousands of frames.
Ability to follow written labeling guidelines exactly, and to ask precise questions when they are silent on a case.
Comfortable working in remote-desktop or browser-based environments.

Nice To Haves

Experience across multiple annotation modalities rather than a single task type.
Experience with aerial, overhead, satellite, or thermal/IR imagery.
Experience with long-form video and tracking work.
Experience reviewing model-assisted pre-labels in a human-in-the-loop pipeline.
Prior work on government or regulated-industry programs.

Responsibilities

Label static imagery and video across a range of sensor types, image qualities, and scene conditions, including both real-world and synthetic data.
Produce 2D bounding boxes, instance and semantic segmentation masks, and keypoint annotations to project specification.
Produce oriented and rotated bounding boxes where scene geometry requires them.
Produce 3D and 6 degrees-of-freedom pose, orientation, and scale annotations on projects that require them.
Produce multi-object tracking annotations, maintaining consistent object identity across full sequences — through occlusion, frame exit and re-entry, scale change, and camera or platform motion.
Review, correct, and accept or reject model-assisted and pre-labeled output; report systematic pre-label failure modes rather than silently correcting the same error frame by frame.
Work strictly to the project ontology and guidelines; escalate ambiguous or out-of-ontology objects rather than guessing.
Meet assigned accuracy and throughput targets, and hold that standard consistently across large batches.
Complete rework promptly from QA feedback, applying the correction to comparable cases in the same batch.
Log edge cases and recurring ambiguities so they can be adjudicated and folded into the guidelines.
Follow all customer data-handling, confidentiality, and information security requirements for the assigned project, and keep required training current.