Ukjae Jeong, 정욱재

Hi, I’m Ukjae Jeong, working as an Machine Learning Engineer, focusing mostly on MLOps and NLP.

Free free to reach out to me.

Experiences

Machine Learning Engineer, Danggeun Market Inc.; Seoul, Korea - December 2020 - Present
Machine Learning Engineer, ScatterLab; Seoul, Korea - December 2019 - November 2020
Developer, ScatterLab; Seoul, Korea - March 2019 - December 2019

ML Google Developer Expert (ML GDE); February 2022 - Present
2nd place, Asean AI/ML AWS Gameday; Seoul, Korea — November 2021 (AWS Blog post link)
Finalist, F8 2019 Hackathon; San Jose, USA — May 2019 (DEVPOST link)
Jpub Beta-reader 7th ~ 10th — July 2018 - August 2021
- Read books to be published in advance and give feedbacks. I have participated in feedback of 10+ books.

I/O Extended 2022 Seoul; Seoul, Korea - June 2022
- Technical Seesion. (“AI/ML at Google I/O 22”, Slide link)
TensorFlow Everywhere Korea; Online - February 2021
- Technical Seesion. (“TensorFlow-Text and RaggedTensor for NLP”, Slide link)
F8 2019 meetup in Seoul; Seoul, Korea — March 2019
- Reviewed the experience of participating in the F8 2019 Hackathon

TensorFlow Org - Issues or PRs
- TFHub profile: https://tfhub.dev/jeongukjae
PyTorch Org - Issues or PRs
- Based on this contributions, I was invited to space-limited events such as PTD2 2020, and PTED 2021.

jeongukjae/nori-clone (NLP)
- Developed standalone Nori (Korean Morphological Analyzer in Apache Lucene) written in C++.
jeongukjae/monkey-compiler (Compiler)
- Implemented a compiler for the Monkey programming language in Go and built a demo web page using WASM.

distilkobert-sentence-encoder (ML)
- Developed sentence embedding model for Korean
- TFHub page: https://tfhub.dev/jeongukjae/distilkobert_sentence_encoder/1
- Blog post: https://jeongukjae.github.io/posts/develop-distilkobert-sentence-encoder/
jeongukjae/tfds-korean (ML)
- Developed a library to use Korean text datasets with simple tensorflow-datasets’ api.
jeongukjae/korean-wikipedia-corpus, jeongukjae/namuwiki-corpus
- Dumped and preprocessed Korean wikipedia text
jeongukjae/mecab-bind (NLP)
- Developed a library to bind MeCab(morphological analyzer, written in C++) to Python3 and TensorFlow custom ops.
jeongukjae/smaller-labse (ML)
- Applied concepts from the paper “Load What You Need: Smaller Versions of Multilingual BERT” and compressed the original LaBSE model.
- TFHub page: https://tfhub.dev/jeongukjae/smaller_LaBSE_15lang/1
jeongukjae/KR-BERT-SimCSE (ML)
- Implemented “SimCSE: Simple Contrastive Learning of Sentence Embeddings” using Korean text data and KR-BERT.
jeongukjae/tta (ML)
- Implemented “Transformer-based Text Auto-encoder” using Korean text data to extract unsupervised sentence embeddings.
jeongukjae/faiss-server (ANN)
- Developed an Approximated Nearest Neighbor server in Golang using Faiss and gRPC-gateway.

jeongukjae/korean-spacing-model (ML)
- Developed a a small ML model to correct whitespaces in Korean. (Demo website)
jeongukjae/regularized-singular-value-decomposition
- Implemented Regularized Singular Value Decomposition in “Improving regularized singular value decomposition for collaborative filtering” using Numpy and MovieLens 10M dataset.
jeongukjae/quasi-rnn
- Implemented quasi-rnn in TensorFlow
jeongukjae/lightgbm-serving
- Developed a lightweight http server for lightGBM.
jeongukjae/CFB
- Developed single header C++ library for Compound File Binary(CFB) Format

jeongukjae/python-mecab
- Developed Python binding for MeCab (Morphological analyzer)
jeongukjae/pytorch-bert
- Developed BERT in pytorch

University of Seoul, March 2017 - Present
- Bachelor of Science in Electrical and Computer Engineering
- Bachelor of Science in Computer Science and Engineering