凉ye L.

Key Skills

Software

LabelImg

CVAT

Top Subject Matter

Large Language Model AI Training

Computer Vision Model Training

Top Data Types

Text

Image

Top Task Types

Question Answering

Object Detection

Freelancer Overview

AI Text Data Annotation Project. Core strengths include LabelImg, CVAT, and ChatGPT. AI-training focus includes data types such as Text and Image and labeling workflows including Question Answering and Object Detection.

Labeling Experience

Computer Vision Image Annotation Project

LabelImgImageObject Detection

I performed comprehensive image annotation tasks including object detection, semantic segmentation, and key point labeling across diverse real-world and industrial scenarios. Responsibilities included conducting secondary quality control, eliminating invalid data, and ensuring annotation reliability for model training. I worked closely with technical teams to resolve annotation issues and adapt templates for evolving model requirements. • Demonstrated expertise in multiple image annotation techniques (bounding box, segmentation, key points). • Enabled higher-quality training data by removing errors and inconsistencies from labeled datasets. • Ensured process alignment with evolving model development needs. • Collaborated with engineers to optimize templates and workflows.

2024 - Present

AI Text Data Annotation Project

LabelImgTextQuestion Answering

I completed text data labeling tasks focused on dialogue, question answering, and instruction-based annotation to support large-scale model training. My work included over 20,000 instances of labeling, validation, and error correction, while ensuring strict adherence to annotation standards. I played a key role in process optimization, improving accuracy and efficiency within the team. • Achieved a labeling pass rate of 99.2% by following rigorous data consistency and accuracy standards. • Carried out continual rule review and contributed to annotator guidelines. • Led initiatives that resulted in a 25% improvement in annotation throughput. • Supported quality control by cross-checking annotated items and correcting issues as needed.

2024 - 2024

Education

Work History