Data Scientist focused on Recommendation Systems & Ranking
- 📊 Experienced in ML pipelines, feature engineering, and large-scale data
processing - 🏆 Strong track record in Kaggle & Domestic AI Competitions
- 🚀 Interested in Real-time personalization & ML system design
Machine Learning Data Mining Recommendation System
- Integrated LightGBM Ranker into the recommendation ranking pipeline for Learning-to-Rank tasks
- Developed comprehensive unit tests for the ranking and recommendation pipeline
- Migrated deep learning models from TensorFlow to PyTorch for improved maintainability and development efficiency
- Developed a review-based restaurant recommendation system using a two-stage recommendation architecture
- Designed and implemented a ranking model for personalized restaurant recommendations
- Built a Python library for user behavior analysis and event-driven data processing
- Designed reusable analytics utilities for behavioral data exploration and feature engineering
- Proposed an enhanced TabNet model for categorical feature learning by incorporating ordered target statistics inspired by CatBoost
- Improved categorical feature representation in deep learning models without relying solely on embedding-based encoding approaches
- Applied CatBoost-style ordered target statistics (Ordered TS) to stabilize training and reduce target leakage in tabular deep learning tasks
- Conducted research on hybrid boosting–deep learning architectures for high-performance tabular data modeling
| Competition | Medal | Rank |
|---|---|---|
| Riiid Answer Correctness Prediction | 🥈 | 78 / 3395 |
| Ventilator Pressure Prediction | 🥉 | 171 / 2605 |
| Indoor Location Navigation | 🥉 | 72 / 1170 |
| Writing Quality Prediction | 🥉 | 187 / 1876 |
| Competition | Medal | Rank |
|---|---|---|
| Dacon Web CTR Prediction | 🏆 | 1 / 772 |
| Credit Card Delinquency Prediction | 🏆 | 2 / 3106 |
| U+ AI Ground | 🏆 | 3 / 658 |
| Toss NEXT ML Challenge | 🏅 | 5 / 709 |