Key Skills

Software

OneForma

Scale AI

Other

Top Subject Matter

No subject matter listed

Top Data Types

Audio

Image

Text

Top Task Types

Prompt Response Writing SFT

Red Teaming

RLHF

Translation Localization

Freelancer Overview

I possess extensive experience across the entire spectrum of data labeling and AI training data construction. I have particular expertise in prompt engineering and model response evaluation, having personally designed and localized various types of prompts, ranging from single-turn to multi-turn interactions. I systematically evaluate model responses against detailed criteria such as accuracy, instruction comprehension, harmfulness, and writing style, contributing to the enhancement of LLM quality and safety. Additionally, by concurrently evaluating responses to image-based prompts and conducting red team activities, I actively participated in enhancing model reliability in real service environments. Based on these achievements, I received high praise from colleagues and took on the role of reviewer, responsible for quality control and providing feedback. 
 Recently, I also participated in a Korean audio transcription project, demonstrating precise work capabilities that maintained an accuracy rate exceeding 96%. Starting with the baseline assignment of 3 tasks, I achieved outstanding results by expanding my allocation to 19 tasks within a short period. This ability to understand diverse data types (text, image, audio) and consistently deliver high-quality results is a key strength that differentiates me in the fields of data labeling and AI training data quality improvement.

Labeling Experience

Transcribing project in Lingotribe

OtherTextTranslation Localization

This project involved listening to a given Korean audio recording and transcribing it accurately according to specified instructions. A unique feature of this project was that the work quota increased if accuracy exceeded 96%, and decreased if it fell below that threshold. Within just a few days of starting the work, my quota rose from the base level of 3 items to 19 items without any deductions, demonstrating my recognized accuracy.

This project involved listening to a given Korean audio recording and transcribing it accurately according to specified instructions. A unique feature of this project was that the work quota increased if accuracy exceeded 96%, and decreased if it fell below that threshold. Within just a few days of starting the work, my quota rose from the base level of 3 items to 19 items without any deductions, demonstrating my recognized accuracy.

2025 - 2025

Transcribing project in Oneforma

OneformaAudioTranslation Localization

This project involved listening to a given Korean audio recording, transcribing it accurately, and evaluating the audio's characteristics. Although I was not much actively involved in the work, my accurate transcription and evaluation skills were recognized, leading to my subsequent role reviewing the work of other contributors.

This project involved listening to a given Korean audio recording, transcribing it accurately, and evaluating the audio's characteristics. Although I was not much actively involved in the work, my accurate transcription and evaluation skills were recognized, leading to my subsequent role reviewing the work of other contributors.

2025 - 2025

Redteaming project in Scale AI

Scale AITextRed Teaming

This project involved classifying and evaluating whether a given prompt was harmful or not, and what type of harm it posed, based on specified criteria. While it wasn't overly complex compared to other tasks, my outstanding work performance was recognized. As a result, I was also assigned the role of reviewing other contributors' work, providing feedback, and performing revisions.

This project involved classifying and evaluating whether a given prompt was harmful or not, and what type of harm it posed, based on specified criteria. While it wasn't overly complex compared to other tasks, my outstanding work performance was recognized. As a result, I was also assigned the role of reviewing other contributors' work, providing feedback, and performing revisions.

2025 - 2025

Image prompt RLHF project in Scale AI

Scale AIImageRLHF

This project involved evaluating model responses to single and multi-turn prompts containing images. I assessed whether the model accurately recognized images and properly understood and responded to the context of previous conversations. This also involved evaluating instruction following, harmfulness, and writing style. While I don't recall the exact duration of this project, it was quite a long-term project, and my outstanding work performance on it was recognized. As a result, I was also assigned the role of reviewing other contributors' work, providing feedback, and overseeing the revision process.

This project involved evaluating model responses to single and multi-turn prompts containing images. I assessed whether the model accurately recognized images and properly understood and responded to the context of previous conversations. This also involved evaluating instruction following, harmfulness, and writing style. While I don't recall the exact duration of this project, it was quite a long-term project, and my outstanding work performance on it was recognized. As a result, I was also assigned the role of reviewing other contributors' work, providing feedback, and overseeing the revision process.

2024 - 2025

RLHF project in Scale AI

Scale AITextRLHF

This project involved evaluating the accuracy, instruction following, harmfulness, and writing style of two responses generated by a model based on a given prompt, ultimately determining which response was superior. It was a long-term project spanning approximately six months, during which my outstanding work performance was recognized. Consequently, I also took on the role of reviewing the work of other contributors, providing feedback, and making revisions.

This project involved evaluating the accuracy, instruction following, harmfulness, and writing style of two responses generated by a model based on a given prompt, ultimately determining which response was superior. It was a long-term project spanning approximately six months, during which my outstanding work performance was recognized. Consequently, I also took on the role of reviewing the work of other contributors, providing feedback, and making revisions.

2024 - 2025

Show more

Education

Work History