Data Scientist · Data Lab, Hanwha General Insurance · Assistant Manager · Seoul, Korea
Personalized decision-making under uncertainty, made causal.
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, across both industry and medical research, through counterfactual reasoning and policy learning rather than pure prediction.
CATE · counterfactual · causal discovery · SCM · semiparametric · partial ID
Identifying what causes what — heterogeneous treatment effects and counterfactuals — with structural causal models, semiparametric estimation, and sensitivity / partial-identification under unobserved confounding.
bandits · RL · OPE · DTR/OTR · policy learning
Turning estimated effects into decisions — optimal policy learning, bandits and reinforcement learning, off-policy evaluation, and dynamic / optimal treatment regimes.
HTE · targeting · recommendation · pricing
The through-line — individualized clinical treatment decisions and industry targeting, recommendation, and pricing as two sides of one methodological core.
An applied case study using only public datasets (LendingClub, iPinYou) that combines CATE estimation for price-sensitivity heterogeneity with SCM-based moderator analysis to design individual-level, risk-based pricing and RTB bidding policies — all findings illustrative and projected, not proprietary.
When real-world data fail strong ignorability, point identification gives way to bounds, proxies, and sensitivity analysis — an honest hierarchy of evidence that connects credible causal claims to semiparametric efficiency.
Customer Segmentation is the unsupervised task of partitioning customers into a finite set of segments by similarity in behavior, value, and preference. A common recipe is latent-factor decomposition followed by clustering: behavioral features → NMF (non-negative, parts-based decomposition) → factor scores → K-Means → segments.
An end-to-end applied case study on the public Dunnhumby dataset — NMF latent factors and K-Means segmentation feeding meta-learner / Causal Forest HTE and an OPE-validated optimal targeting policy, with a candid look at positivity violation and counter-intuitive "sleeping dog" segments.
One methodological spine — estimate heterogeneous treatment effects and turn them into individual-level policies — powers both clinical sequential treatment decisions and industrial targeting, pricing, and recommendation.
A case study on extracting a 3-layer attribute taxonomy (product / perceptual / theory-grounded) with LLM/VLM pipelines, turning it into user profiles and a mixture-of-experts adaptor, and plugging it into standard recommenders across two public domains (fashion + music).
PPFL: A personalized progressive federated learning method for leveraging different healthcare institution-specific features
Wicox: Weight-based integrated Cox model for time-to-event data in distributed databases without data-sharing
Hanwha General Insurance (former Carrot General Insurance)
Yonsei University · Biomedical Systems Informatics Lab
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
Cheongju University · Cheongju, Korea
GPA 4.24 / 4.50 (top of class)
taehyun9573@gmail.com ·Seoul, Republic of Korea