Lee Y.

Key Skills

Software

No software listed

Top Subject Matter

LLM training data annotation and text classification

Vision dataset annotation (boxes/polygons/segmentation/tracking)

Top Data Types

Text

Image

Top Task Types

Fine-tuning

Segmentation

Freelancer Overview

Text Annotation for LLM Training (intent/sentiment/dialogue fine-tuning & compliance review). Core strengths include Internal and Proprietary Tooling. AI-training focus includes data types such as Text and Image and labeling workflows including Fine-tuning and Segmentation.

Labeling Experience

Image Annotation Specialist (vision labeling & QC to SOP)

ImageSegmentation

Completed end-to-end image annotation workflows including bounding box, polygon, key point, semantic segmentation, and video frame tracking. Performed image data cleaning, screened invalid samples, and applied privacy data desensitization to meet annotation SOPs. Ensured accurate boundary control with pixel-level and classification standards while preventing common errors such as missing annotations, incorrect labels, and bounding box deviation. • Bounding box, polygon, and key point annotation • Semantic segmentation and video frame tracking • Image cleaning, invalid screening, and privacy desensitization • Quality control for boundary accuracy and labeling consistency

Present

Text Annotation for LLM Training (intent/sentiment/dialogue fine-tuning & compliance review)

TextFine-tuning

Performed LLM training data preprocessing including text cleaning, deduplication, error correction, invalid sample filtering, and semantic standardization. Conducted text classification and user intent recognition with sentiment annotation to support downstream model training and dialogue fine-tuning. Reviewed non-compliant content and applied context-aware judgment to distinguish similar labels, resolve ambiguous boundaries, and catch verbal sarcasm or other semantic traps. • LLM training text cleaning and deduplication • Intent, sentiment, and dialogue/QA related labeling • Non-compliant content review and invalid sample filtering • Semantic consistency checks to reduce label confusion

Present

Education

Work History