Data Science and Innovation student at Thammasat University, available for a co-op placement from January to April 2027. I work with data engineering, statistical analysis, machine learning, and practical analytics tools.
- Building reviewable data workflows with Python, SQL, DuckDB, and Parquet
- Interested in analytics, learning-to-rank research, data governance, and model evaluation
- Tools: Excel, Power BI, Tableau, and Git
- Studying cybersecurity alongside Data Science and Innovation
Methods include data modeling, ETL, data quality checks, data contracts, hypothesis testing, direct standardization, multiple-testing correction, classification, learning to rank, text embeddings, fairness audits, and model evaluation.
At UMEDA Co., Ltd. in Bangkok, I built Excel workbooks that broke down pharmaceutical sales by manufacturer and product across government, private, and OTC channels, grouped by ATC3 therapeutic class. I compared hospital and reference prices, used Excel trend and forecast tools, and prepared Power BI dashboards and PowerPoint decks for executive meetings.
Academic project for the Ministry of Justice's Kamlangjai program. I linked 12 corrections datamart tables with probation data in DuckDB into person-level Parquet tables covering 2M+ people, with a data contract that identifies usable, proxy, and protected features. The work compares occupational-grant recipients with non-recipients using direct standardization and Benjamini–Hochberg correction, with aggregate-only reporting. Classification models are framed as a rule-consistency check: leading features were inputs to the official classification rule, so they are not evidence of an independent predictive signal.
Python SQL DuckDB Parquet scikit-learn
Independent research project on ranking MITRE ATT&CK techniques for CVE roles. The public repository’s main branch contains the earlier W3 foundation; the research results are pinned at 2a2a6cf on work/w8-ablations. Reported sealed-test results across 158 queries: hybrid nDCG@5 0.5581, text-only LambdaRank 0.5317, MiniLM 0.2178, and BM25 0.1451. Ablations attribute the hybrid gain to platform (CPE) features rather than graph structure; a randomized graph scored comparably to the real graph.
Python LightGBM LambdaRank BM25 MiniLM pandas
Academic research project using 20,919 undergraduate records (2014–2024). I cleaned course-level records, built credit-weighted term GPAs and academic features, and compared logistic regression, random forest, and XGBoost at three checkpoints. The reported XGBoost ROC-AUC increased from 0.866 after Term 1 to 0.992 after Term 3.
Python pandas scikit-learn XGBoost
ArguSQL is a Windows application for querying local CSV, TSV, Parquet, JSON, and JSONL files with DuckDB. It includes live preview, autocomplete, a visual JOIN builder, and CSV export. Its 21 September 2026 README reports 264 native tests and separate 9/9 packaged workflow checks.
TypeScript Electron React DuckDB
Thammasat University · 2023–2027 (expected)
B.Sc. in Data Science and Innovation (Cybersecurity module), College of Interdisciplinary Studies
GPA 3.33 / 4.00; last four semesters 3.56
A Machine Learning-Based Student Dropout Risk Prediction and Early Warning System
Second author · accepted in Asian Health, Science and Technology Reports, 34(4), 2026 · presented at AJCC 2026, Sapporo, Japan
