Year 3 Analytical Testing Science student at Hong Kong Metropolitan University.
I build small research and decision-support systems with a simple rule: if the available evidence cannot support a comparison or conclusion, the system should say so rather than manufacture certainty.
- ReasonedOps — service-operations prototype for routing, audit trails and guarded outcome evaluation; can withhold a department comparison when the observed design cannot support it.
- Uni Life Planner — R Rebuild — reconstruction of a Year 1 text-analytics project into a reproducible dictionary baseline with abstention; no external accuracy claim until independently labelled data exists.
- Sector Rotation Forecasting — audit and reconstruction of a 2025 forecasting prototype; a structural-null test showed that the original coordinate-forecasting headline was not sufficient evidence of market predictability.
Python · R · statsmodels · scikit-learn · FastAPI · GitHub Actions · reproducible statistical workflows