Heewon Jeon

Heewon is an applied scientist at SK Telecom, where he solves many MNO (mobile network operator) business problems with data, including churn prediction with user activities and credit scoring for effective bond collection. Previously, he worked as a search engineer at Yahoo!. He has almost 16 years of data science experience, specializing in statistical modeling, data analysis, and natural language processing.

He enjoys developing NLP open source projects as a hobby, and one of his activities is contributing to GluonNLP as a member of the Distributed (Deep) Machine Learning Community, or DMLC. He likes to use MXNet as his main deep learning platform because of the efficiency of training. With MXNet and GluonNLP, to improve Korean NLP performances, he trained a large scale Korean BERT model (aka KoBERT). Recently, he successfully trained Korean GPT2 with hundreds of millions of sentences on multiple machines by cooperating with AWS internal teams. He is also an author on the MXNet and the AWS Korea blog, and has written numerous articles on model training and distribution.