Current Status

Things that I am being focused on

Auto Data Wrangling

In our definition, Auto Data Wrangling (AutoDW) is an automation of the process to make tabular/timeseries data ready for downstream tasks (e.g. machine learning, data analytics). AutoDW involves various stages: table integration, prediction engineering, advanced data enrichment & cleaning. Since there are no silver bullet solutions for the complicated task, I have worked on the research by mixing several technologies: classical machine learning models (e.g. Random Forest), pre-trained language models (e.g. BERT), and large language models (e.g. GPT-4).

AutoML

Although AutoML for tabular/timeseries data has become established with the use of exploratory methods or meta-learning methods, they lack generalizability to various datasets. I believe that foundation models for tabular data overcome the challenging. Since there are no standards of foundation models for tabular data, I am working on building the self-supervised model for tabular data.

Explore US

I really love exploring big-scale nature in US since I was inspired by Yellowstone National Park in the first visit. California is the one of the best places to access nature.

Recent Favorite Place

Pinnacles National Park