About · Tae Hyun Kim (Lowell)

I am a data scientist whose research centers on causal inference and data-driven decision-making for high-stakes, individualized recommendations. To that end I draw on statistical and probabilistic modeling, optimization, ML/AI, and LLM-based agentic systems — approaching applied problems through counterfactual reasoning and policy learning rather than pure prediction.

The through-line is personalization: individualized clinical treatment decisions and industry targeting are, to me, two sides of one methodological core — which is why I move between medical research and the insurance industry.

This site

Study and research notes, published with Astro from an Obsidian vault — only notes explicitly marked publish: true. Networked by wikilinks and backlinks; code, data, and experiments live in a local workspace, while concepts and connections live here.

Browse notes → GitHub Email

Publications

PPFL: A personalized progressive federated learning method for leveraging different healthcare institution-specific features

Kim, T. H., Yu, J. Y., Jang, W. S., Heo, S. C., Sung, M., Hong, J., … Park, Y. R. (2024). iScience, 27(10).First author

Wicox: Weight-based integrated Cox model for time-to-event data in distributed databases without data-sharing

Park, J. A., Kim, T. H., Kim, J., & Park, Y. R. (2022). IEEE Journal of Biomedical and Health Informatics, 27(1), 526–537.Co-author

Patents

Method and apparatus for analyzing safe driving behavior of vehicle drivers — KR 10-2024-0115693 (2024)
Personalized federated learning method and program for implementing the same — KR 10-2021-0097839 (2021)

Experience

Data Scientist · Data Lab

Hanwha General Insurance (former Carrot General Insurance)

Lead — Multi-agent LLM system for insurance sales operations: Supervisor + 6 specialist agents, on-prem sLLM, a Metric Registry over ~50 KPIs / 27 tables with auditable provenance (no free-form SQL).
Lead — Causal-inference module of an end-to-end marketing optimization platform: causal discovery & SCM over the funnel, meta-learner CATE per campaign, CLV-based optimal policy learning.
Lead — Multi-touch attribution & cost-aware budget optimization: inhomogeneous Poisson process with right-censoring, Incremental Shapley credit assignment, marginal-efficiency re-allocation.
Lead — Reward-marketing CATE (R-Learner) + segmentation targeting → +13% company-wide conversion uplift (+45% in the top segment).

Mar 2022 – Present

Graduate Research Assistant

Yonsei University · Biomedical Systems Informatics Lab

Personalized progressive federated learning (PPFL) clinically validated across five hospitals (breast/liver/colorectal/lung cohorts) — distributionally robust learning across institutions.
Multi-institutional CDM data mart; propensity-score matching & multi-site meta-analysis of immune-checkpoint-inhibitor adverse reactions; uncertainty-quantified, explainable time-series models.

Mar 2020 – Feb 2023

Education

M.S. in Biomedical Systems Informatics

Yonsei University, College of Medicine · Seoul, Korea

Advisor: Prof. Yu Rang Park · GPA 4.03 / 4.30 · Thesis: Personalized Progressive Federated Learning with Client-Specific Vertical Features

Mar 2020 – Feb 2023

B.S. in Industrial Engineering

Cheongju University · Cheongju, Korea

GPA 4.24 / 4.50 (top of class)

Mar 2014 – Feb 2020