face_img

HARUYA UMEMOTO

ML Engineer

Biography

Haruya Umemoto studied computer science at university/graduate school and conducted research using machine learning and deep learning. His research interests include reinforcement learning and natural language processing (NLP), and he has recently developed an understanding of research in the field of imaging. After graduating, he joined an AI venture, where he was in charge of research and development of large-scale language models, dialogue systems, etc., with a focus on NLP. In his current position, he focuses on providing AI solutions using his own knowledge, and is also involved in setting up other AI teams and product development. He is interested not only in the ML area but also in IT in general, and has skills in front-end, back-end and cloud computing.

NEWS

Recently, I have been supporting the launch of generative AI projects and consulting. Please contact me from the Contact in the upper right corner. (If there is no response, it's possible that it might have been overlooked, so please contact me via X or LinkedIn.)

Skills

Language: Python, Dart, TypeScript, Rust, Go, C++, Java, JavaScript

クラウド: Google Cloud(LOVE), Firebase, AWS, Azure

Framework: PyTorch, Docker, Flutter, React, Next.js, Terraform

Specialization: ML, DL, RL, NLP, Generative AI, Data analysis, Dialogue system

Prizes/Certifications

2023 - Ledge.ai CHALLENGE Generative AI Hackason Outstanding Award 2022 - GCP Professional Data Engineer 2020 - JSAI Outstanding Paper Award 2020 - Komoda Advanced Science Academic Award 2020 - The Highest Grade Point Award 2020 - Outstanding Presentation of Master Thesis Award 2017 - Data analysis&Simulation Hackason 2017 Second prise (SIG-DOCMAS) 2015 - Award of Aoyama gakuin University (Hack U 2015 at Aoyama gakuin University)

Career

2021/10, ML Engineer, TC3 2020/04, ML Engineer, Arithmer 2020/03, Graduated shcool of Aoyama gakuin university, Master of engineering 2018/08, Internship, Cookpad 2018/03, Aoyama gakuin University, Bachelor of engineering

Pubilications

学習の安定化のために方策の埋め込みを利用する強化学習手法の検討

梅本晴弥,豊田哲也,大原剛三:学習の安定化のために方策の埋め込みを利用する強化学習手法の検討,知識ベースシステム研究会(SIG-KBS) (2019).

食材名の分散表現学習を用いた料理レシピの栄養推定手法

梅本晴弥,豊田哲也,大原剛三:食材名の分散表現学習を用いた料理レシピの栄養推定手法,行動変容と社会システム vol.05 (2019).

k-NN Based Forecast of Short-Term Foreign Exchange Rates

Umemoto, Haruya, Tetsuya Toyota, and Kouzou Ohara. “k-NN Based Forecast of Short-Term Foreign Exchange Rates.“ Pacific Rim Knowledge Acquisition Workshop. Springer, Cham, 2018.

料理レシピの分散表現を用いた代替食材の発見手法

梅本晴弥,豊田哲也,大原剛三:料理レシピの分散表現を用いた代替食材の発見手法,行動変容と社会システム vol.03 (2018).

過去の変動に対する類似検索を用いた短時間USD/JPY為替レート予測

梅本晴弥,豊田哲也,大原剛三:過去の変動に対する類似検索を用いた短時間USD/JPY為替レート予測,研究報告知能システム(ICS),2017-ICS-186,pp. 1-7 (2017).

Slides

執筆した記事

Copyright © 2023, Haruya Umemoto