박장원 Park
Machine Learning Engineer at BHSN
Seoul, South Korea
Summary
👤
Senior🎓
Top School박장원 Park is a Machine Learning Engineer based in Seoul with nine years of experience specializing in Natural Language Processing and large language models. He has applied Korean-focused pretrained models in production and research settings, notably contributing to the popular KoELECTRA project by finetuning, expanding task support (KorSTS, QuestionPair, Korean Hate Speech) and hardening training/evaluation pipelines. His background spans R&D and industry roles at Samsung Research, Mathpresso, KB Kookmin Bank, and BHSN, blending research-grade modeling with pragmatic engineering for deployed systems. Equally at home in data engineering and model training, he has implemented robust data loaders and training logic for JointBERT-style intent/slot systems. A Yonsei alumnus with dual interests in business and computer science, he brings a product-minded approach to NLP research and open-source collaboration.
9 years of coding experience
3 years of employment as a software developer
Bachelor's degree, Business Administration & Computer Science, 4.1 / 4.3, Bachelor's degree, Business Administration & Computer Science, 4.1 / 4.3 at Yonsei University
Korean, English
Github Skills (15)
sequence-labeling10
transformers10
data-loading10
pytorch10
machine-learning10
nlp10
language-model10
python10
electra10
bert10
fine-tuning10
classification10
data-processing10
question-answering7
tensorflow4
Programming languages (6)
Github contributions (5)
monologg/KoELECTRA
Apr 2020 - Jan 2022
Pretrained ELECTRA Model for Korean
Role in this project:
ML Engineer
ML EngineerContributions:3 reviews, 49 commits, 11 PRs in 1 year 9 months
Contributions summary:박장원's contributions primarily revolve around finetuning and adapting the KoELECTRA model for various Korean NLP tasks. They added support for tasks like KorSTS and QuestionPair, along with Korean Hate Speech. Significant changes involve modifications to training and evaluation scripts, including updates for compatibility with different transformer versions and the introduction of new model architectures. The user also addresses bug fixes, such as correcting typos in the optimizer saving process and refining logging mechanisms.
nlppytorchlanguage-modeltransformersbert
monologg/JointBERT
Nov 2019 - Jun 2021
Pytorch implementation of JointBERT: "BERT for Joint Intent Classification and Slot Filling"
Role in this project:Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:62 commits, 2 PRs, 51 pushes in 1 year 7 months
Contributions summary:박장원 appears to have primarily focused on the development of the `data_loader.py` file, implementing core components for data loading and feature conversion. Their contributions involve creating input examples and features tailored for the model, specifically for a SemEval dataset, and later expanding it for JointBERT models like ATIS and SNIPS, reflecting their involvement in machine learning model development and data processing. The user also made changes to other key files like `trainer.py`, which indicates they worked on the training process for the model, including evaluation metrics and prediction.
pytorchnlptransformersbertintent-classification
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.