Key Skills

Software

Labelbox

Label Studio

Top Subject Matter

No subject matter listed

Top Data Types

Computer Code Programming

Image

Text

Top Task Types

Classification

Computer Programming/Coding

Data Collection

Object Detection

Translation/Localization

Freelancer Overview

AI Training Data and Pipeline Competencies I possess strong hands-on experience in deep learning model training and data pipeline development, gained through intensive AI bootcamps and various projects. My key strength lies in the implementation of MLOps pipelines using LabelStudio, where I successfully integrated data crawling, preprocessing, auto-labeling, and Active Learning to enhance data quality and labeling efficiency. Furthermore, I have developed a real-world healthcare service utilizing Next.js, FastAPI, and LangChain, demonstrating my ability to understand and contribute to the full cycle of AI product development. With practical experience in models like YOLO, ViT, and LSTM, combined with Big Data Analysis, ADsP, and SQLD certifications, I am well-equipped to quickly adapt to complex guidelines and deliver high-quality training data that meets specific model requirements.

Labeling Experience

YouTube Thumbnail Emotion and View Count Analysis Project

OtherImageBounding BoxClassification

Data Collection & Curation: Used the yt-dlp library to scrape video metadata and images. Ensured data quality by performing manual curation to select only faces with clearly expressed emotions for training. * Deep Learning Implementation: Utilized a pre-trained YOLO model for robust face detection, followed by Transfer Learning on a Vision Transformer (ViT) to classify 7 facial emotions (approx. 70% accuracy). * Statistical Analysis: Applied advanced Python libraries to perform statistical correlation analysis between the recognized facial emotions and video engagement metrics (view count). This process demonstrated proficiency in integrating complex computer vision results with tabular data analysis to derive project conclusions.

Data Collection & Curation: Used the yt-dlp library to scrape video metadata and images. Ensured data quality by performing manual curation to select only faces with clearly expressed emotions for training. * Deep Learning Implementation: Utilized a pre-trained YOLO model for robust face detection, followed by Transfer Learning on a Vision Transformer (ViT) to classify 7 facial emotions (approx. 70% accuracy). * Statistical Analysis: Applied advanced Python libraries to perform statistical correlation analysis between the recognized facial emotions and video engagement metrics (view count). This process demonstrated proficiency in integrating complex computer vision results with tabular data analysis to derive project conclusions.

2025 - 2025

Korean Audio Transcription

LabelboxTextTranslation Localization

Quality-Driven Transcription: Performed large-scale Korean audio transcription using the industry-standard Labelbox platform for ASR training data. Achieved a high average Accuracy Rate of 93% following external Quality Control (QC) review, demonstrating exceptional quality assurance and compliance with rigorous guidelines. * Technical Accuracy: Conducted detailed QA including Speaker Diarization, non-verbal expression handling, and standardization of complex Korean orthography to minimize the Word Error Rate (WER). This ensured the delivery of high-quality, model-ready training data under tight deadlines.

Quality-Driven Transcription: Performed large-scale Korean audio transcription using the industry-standard Labelbox platform for ASR training data. Achieved a high average Accuracy Rate of 93% following external Quality Control (QC) review, demonstrating exceptional quality assurance and compliance with rigorous guidelines. * Technical Accuracy: Conducted detailed QA including Speaker Diarization, non-verbal expression handling, and standardization of complex Korean orthography to minimize the Word Error Rate (WER). This ensured the delivery of high-quality, model-ready training data under tight deadlines.

2025 - 2025

Food Detection Modeling, Service Development

Label StudioImageBounding BoxClassification

1. MLOps Pipeline Implementation and Efficiency Maximization - Constructed an end-to-end MLOps pipeline using LabelStudio to automate food image data crawling, preprocessing, and labeling. - Implemented Active Learning to prioritize the labeling of high-uncertainty samples, optimizing labeling costs and time. - Developed automated features for synchronizing labeler updates and downloading data, ensuring efficiency in the data acquisition/processing flow. 2. Image Analysis and Service Integration - Utilized a YOLO model for accurate detection and recognition of various ingredients and food items from user-uploaded images. - Built a fast and stable backend API using FastAPI to deliver food data analysis results to the frontend. 3. AI-Powered Personalized Service Delivery - Developed the user-friendly web interface using Next.js. - Employed LangChain to generate and provide personalized healthcare feedback, such as nutritional information and diet guidance

1. MLOps Pipeline Implementation and Efficiency Maximization - Constructed an end-to-end MLOps pipeline using LabelStudio to automate food image data crawling, preprocessing, and labeling. - Implemented Active Learning to prioritize the labeling of high-uncertainty samples, optimizing labeling costs and time. - Developed automated features for synchronizing labeler updates and downloading data, ensuring efficiency in the data acquisition/processing flow. 2. Image Analysis and Service Integration - Utilized a YOLO model for accurate detection and recognition of various ingredients and food items from user-uploaded images. - Built a fast and stable backend API using FastAPI to deliver food data analysis results to the frontend. 3. AI-Powered Personalized Service Delivery - Developed the user-friendly web interface using Next.js. - Employed LangChain to generate and provide personalized healthcare feedback, such as nutritional information and diet guidance

2025 - 2025

Education

Work History