Aditya Kumar

Key Skills

Software

Label Studio

Other

CVAT

Top Subject Matter

Computer Vision Object Detection

Top Data Types

Video

Image

Text

Document

Top Task Types

Segmentation

Classification

Action Recognition

Bounding Box

Freelancer Overview

Personal Computer Vision Annotation Project. Core strengths include CVAT. AI-training focus includes data types such as Image and labeling workflows including Bounding Box.

Labeling Experience

Video Annotation Practice Project (Self-Learning)

Label StudioVideoSegmentationAction Recognition

In this self-learning project, I performed video annotation for AI training using Label Studio. I focused on identifying relevant moments and accurately marking temporal boundaries. Actions and visible objects were classified according to instructions. Metadata was added to enhance dataset usability. • Annotated short videos for AI training. • Identified and marked accurate temporal boundaries. • Classified actions and objects based on instructions. • Tagged metadata to improve dataset quality.

In this self-learning project, I performed video annotation for AI training using Label Studio. I focused on identifying relevant moments and accurately marking temporal boundaries. Actions and visible objects were classified according to instructions. Metadata was added to enhance dataset usability. • Annotated short videos for AI training. • Identified and marked accurate temporal boundaries. • Classified actions and objects based on instructions. • Tagged metadata to improve dataset quality.

2022 - Present

Image Annotation Practice Project

CVATImageClassificationBounding Box

In this project, I independently constructed an end-to-end dataset for object detection using CVAT and Python. I applied bounding boxes, polygons, and keypoint labels for over 500 images, and enforced rigorous standards for label consistency and inter-annotator agreement. Self-QA reviews were performed to minimize errors and simulate professional annotation workflows. • Developed structured annotations for object detection and segmentation tasks • Utilized CVAT and Python to label images with bounding boxes, polygons, and keypoints • Emphasized quality control through self-review to reduce errors • Replicated throughput and accuracy expected in production-level annotation pipelines.

In this project, I independently constructed an end-to-end dataset for object detection using CVAT and Python. I applied bounding boxes, polygons, and keypoint labels for over 500 images, and enforced rigorous standards for label consistency and inter-annotator agreement. Self-QA reviews were performed to minimize errors and simulate professional annotation workflows. • Developed structured annotations for object detection and segmentation tasks • Utilized CVAT and Python to label images with bounding boxes, polygons, and keypoints • Emphasized quality control through self-review to reduce errors • Replicated throughput and accuracy expected in production-level annotation pipelines.

2022

AI Trainer (Aether Generalist) | Outlier AI | Freelance, Remote

OtherImage

As an AI Trainer (Aether Generalist) at Outlier AI, I evaluated AI-generated responses for accuracy, clarity, and quality as part of large-scale model training. My responsibilities included image and video data annotation, maintaining high standards and meeting deadlines. I consistently delivered high-quality feedback to improve model outputs. • Conducted LLM response evaluation with 95% accuracy. • Performed data annotation in the image and video domains with a 90% success rate. • Worked approximately 7–8 hours per week in a freelance, remote capacity. • Ensured thorough evaluation and review of AI-generated outputs.

As an AI Trainer (Aether Generalist) at Outlier AI, I evaluated AI-generated responses for accuracy, clarity, and quality as part of large-scale model training. My responsibilities included image and video data annotation, maintaining high standards and meeting deadlines. I consistently delivered high-quality feedback to improve model outputs. • Conducted LLM response evaluation with 95% accuracy. • Performed data annotation in the image and video domains with a 90% success rate. • Worked approximately 7–8 hours per week in a freelance, remote capacity. • Ensured thorough evaluation and review of AI-generated outputs.

2025 - 2025

Education

Work History