Kaichi Irie (入江海地)
日本語System Optimization Laboratory, Applied Mathematics and Physics Course, Graduate School of Informatics, Kyoto University
I'm a second-year master's student at the Graduate School of Informatics, Kyoto University, researching continuous optimization — particularly black-box optimization. I also contribute to Optuna, an open-source hyperparameter optimization framework.
Research
Optuna Constrained Tree-Structured Parzen Estimator Is a Joint Density Generalization of c-TPE
Batch Acquisition Function Evaluations and Decouple Optimizer Updates for Faster Bayesian Optimization
Open Source
Hyperparameter optimization framework. Mostly during my internship at Preferred Networks: speeding up the GPSampler, and documentation work.
Internships
Recruit Co., Ltd. Software Engineer Intern
Preferred Networks Software Engineer Intern
Worked on speeding up the GPSampler in Optuna.
Nikkei Software Engineer Intern
Designed and built a preprocessing pipeline for LLM training data on Amazon SageMaker.
AbemaTV (ABEMA) Data Scientist Intern
Projects
A CPU, assembler and compiler built up from NAND gates, written in Go. The Jack compiler runs in the browser.
Education
Graduate School of Informatics, Kyoto University M.S. in Informatics
System Optimization Laboratory, Applied Mathematics and Physics Course(opens in a new tab)
Faculty of Economics, Kyoto University B.A. in Economics
Skills
- Languages
- Python, Go, Julia, C++
- Focus
- Continuous optimization, black-box optimization, Bayesian optimization, hyperparameter optimization
- Tools
- Optuna, PyTorch, Amazon SageMaker, Git, Docker
Languages
- Japanese
- Native
- English
- Business level
Certifications
- Japan Statistical Society Certificate, Grade 1 (Award for excellence in Statistical Mathematics) Certificate(opens in a new tab)Oct 2024
- TOEIC L&R 940 Certificate(opens in a new tab)Apr 2024
- Bookkeeping Certificate, Level 2 (JCCI)Nov 2021