Oleksandr Shtunder

Key Skills

Software

CVAT

Labelbox

Roboflow

Top Subject Matter

No subject matter listed

Top Data Types

3D Sensor

Image

Video

Top Label Types

Bounding Box

Polygon

Segmentation

Freelancer Overview

I am a detail-oriented AI Data Labeling Specialist with one year of freelance experience building and refining datasets for computer vision models. I specialize in complex spatial data annotation, with a strong focus on annotating architectural features, interior layouts, and structural elements. My background allows me to accurately navigate multiple design variations and maintain strict spatial logic when labeling 3D and 2D environments, giving me a unique edge in structural object detection. I am highly proficient in pixel-perfect semantic segmentation, 3D cuboids, and precise bounding box annotations using tools like CVAT and Labelbox. Consistently maintaining a 98%+ accuracy rate on NDA-protected projects, I excel at identifying ambiguous edge cases, performing thorough self-QA, and delivering high-quality, high-volume tasks independently.

Labeling Experience

Interior Spatial Mapping & Structural Annotation

CVATImageBounding BoxPolygon

Annotated high-resolution 2D and 3D visual data of interior spaces to train a spatial mapping and structural recognition computer vision model. The core tasks involved pixel-perfect semantic segmentation of architectural elements, furniture, and structural boundaries across various design layouts and building variations. Processed a high-volume dataset of over 15,000 images/frames while strictly adhering to complex spatial logic guidelines. Consistently maintained a 98.5% Quality Assurance (QA) score by performing rigorous self-checks before submitting batches.

Annotated high-resolution 2D and 3D visual data of interior spaces to train a spatial mapping and structural recognition computer vision model. The core tasks involved pixel-perfect semantic segmentation of architectural elements, furniture, and structural boundaries across various design layouts and building variations. Processed a high-volume dataset of over 15,000 images/frames while strictly adhering to complex spatial logic guidelines. Consistently maintained a 98.5% Quality Assurance (QA) score by performing rigorous self-checks before submitting batches.

2025

Education

Work History