Yejin Hwang Yejin-Hwang

👋 Hi, I'm Yejin (예진 황)

🎓 MS in Data Science @ Texas A&M University–Corpus Christi (GPA 4.0/4.0 | Graduate Fellow)
🎓 B.S. in Culture & Technology + Biotechnology @ Sungkyunkwan University (SKKU)
🔬 Former Data Science Researcher @ Samsung Medical Center (Epidemiology & Clinical Data)
📊 Passionate about financial time-series forecasting, NLP, and turning messy real-world data into actionable insights

📄 KDD 2026 (ACM SIGKDD) — Research paper under review on multimodal financial forecasting
🔗 Financial Time-Series Forecasting with Deep Learning and Social Media Sentiment
🥇 Awarded the SKKU President Award (1st Place, R&D Innovation) for a methane-reducing functional gum project (₩2M prize, graduation evaluation waived)
🔗 View SKKU article (English)
🎤 Presented stock forecasting research at Coastal Bend Conference (2025) and NMSU Workshop (2025)
🔗 View LinkedIn post

Graduate Research Assistant @ Prof. Sreelekha Guggilam's Lab (TAMUCC) · 2024.09 – Present

Thesis research at the intersection of financial time-series modeling and social sentiment analysis
Designed a Reddit sentiment-enhanced forecasting framework using FinBERT + multivariate transformer models (TFT, Chronos-T5, TimesFM)
Fine-tuning large foundation models on stock price data; building scalable inference pipelines for real-time market prediction

Clinical Data Research Assistant @ Samsung Medical Center – CCE · 2022.08 – 2023.08

Supported data collection and analysis for clinical epidemiological studies under Prof. Juhee Cho (SAHIST / Johns Hopkins) and Prof. Danbee Kang (SAHIST)
Worked on structured health datasets related to rare and oncological diseases
Managed and cleaned clinical datasets; assisted in data coordination (protocols, data workflows) for epidemiology research

Research Assistant @ Prof. Jinhee Hur's Lab (SKKU) · 2022.03 – 2022.07

Conducted scientific literature review and basic data handling for lab-scale experiments under Prof. Jinhee Hur (Ph.D. from Johns Hopkins, Postdoctoral Fellow at Harvard)
Gained foundational experience in academic research and interdisciplinary collaboration

[KDD 2026 — Under Review] Financial Time-Series Forecasting with Deep Learning and Social Media Sentiment
🔗 Repo

Full-scale multimodal forecasting framework integrating Reddit-derived FinBERT sentiment + LOESS spike features into a Temporal Fusion Transformer (TFT) across 10 large-cap U.S. equities and 1,000 trading days
Achieved +213.6% average IC improvement over TFT baseline; best model reached 20.19% cumulative return and Sharpe ratio of 2.74 in the test period
Expanded and formalized from Reddit-Driven-Stock-Forecasting for KDD 2026 submission

Clean, reproducible pipelines for forecasting stock prices (TSLA, NVDA) using ARIMA, Google TimesFM, Amazon Chronos, and TFT with Reddit sentiment and spike features
TFT + Reddit achieved RMSE ↓ 40.2% on TSLA and ↓ 87.9% on NVDA vs. baseline TFT

Compared Bayesian and Frequentist linear regression on individual medical cost prediction using the rethinking package in R
Focused on uncertainty quantification and interpretability