Data Science undergraduate at Universitas Brawijaya (Malang, Indonesia). I build forecasting, computer vision, NLP and analytics projects, and I enter national data science competitions.
Portfolio: vincent-osbert.github.io · LinkedIn: vincent-osbert · Email: vosbert151515@gmail.com
| Result | Competition | Task |
|---|---|---|
| 8th of 260 teams (private LB) | ARA ITS 7.0 | Pothole segmentation, SegFormer MiT-B5, Dice 0.804 · repo |
| 9th of 37 teams (private LB) | National Statistics Challenge 2026 | Air humidity forecasting from IoT sensors, 4-model ensemble with chronological CV, RMSE 3.91 · repo |
| 12th of 13 teams (private LB) | FIK FAIR 2026 | Tabular regression on 382 rows, kernel hybrids + TabPFN, row-order leak found and not used · repo |
| National finalist | DataQuest 5.0, Universitas Airlangga | Life-satisfaction prediction for 95,609 survey respondents |
- olist-delivery-review-diagnosis: SQL root-cause analysis; late deliveries explain 63% of a review drop.
- hotel-cancellation-overbooking: late-cancellation model turned into a newsvendor overbooking rule.
- retail-demand-inventory-planner: LightGBM demand forecast to stock target, 10.83% test WAPE vs 16.60% baseline.
- telco-churn-retention: churn ranking, 74.8% precision in the top 20% risk group.
- freelance-data-projects: Excel data cleaning and Indonesian text annotation QA.
Python · SQL · pandas · scikit-learn · LightGBM · PyTorch · Hugging Face · Streamlit