I work across data science, analytics engineering, machine learning validation, and risk analytics, with experience spanning healthcare modeling and large-scale operational systems.
My recent technical work focuses on model evaluation, calibration, temporal validation, local model updating, reproducible analysis, and data-quality controls. Earlier industry work at Ping An Finserve focused on structured risk operations, strategy systems, analytics workflows, and product management across nationwide financial portfolios.
Python · SQL · R · scikit-learn · LightGBM · Model Validation · Calibration · Data Modeling · Risk Analytics · Reproducible Pipelines
External validation · calibration · local model updating · robustness · temporal validation
Public-safe evidence package and runnable reconstruction of breast cancer risk-model validation work originating from my BC Cancer / UBC MDS capstone, with later independent extensions. The retrospective analysis covered up to 438,571 eligible screening exams across 1–5 year cumulative risk horizons and included discrimination, calibration, patient-cluster bootstrap, subgroup analysis, first-exam sensitivity analysis, leakage auditing, decision-curve analysis, and future-period temporal validation.
Causal diagnosis · randomized policy evaluation · risk economics · adaptive decisions
An independent simulated case study investigating a fraud loss-rate increase from 0.32% to 0.47%. I follow the evidence from payment routing and a randomized authentication intervention to cross-channel displacement and recovery congestion. The runnable analysis compares policy value, including cases where adaptive rules tie or underperform fixed policies.
Explore the case study · Open Policy Lab
Product management · nationwide debt-collection operations · structured workflows · analytics systems
As a Product Manager in the Risk Asset Management Center, I worked on CMP2.0 and Tianshu, supporting collection operations across credit-card, consumer-loan, retail-finance, and mortgage portfolios. The work covered spreadsheet-to-system migration, automated operational logging, differentiated workflows across business lines, intelligent dialing strategies, analytics dashboards, strategy middleware, tag and permission models, and execution channels including calls and SMS.
These projects provide the industry and operational context behind my current work in data science: how analytical outputs connect with data structures, workflows, decision logic, and real operating constraints.
Healthcare & ML validation — external validation, calibration, model updating, robustness, temporal validation
Risk & operational analytics — structured workflows, strategy systems, analytics dashboards, financial operations
Platform product systems — earlier work at Qunar and Lilith Games across booking, administration, and internal operational tooling
|
I’m writing a fantasy novel in Chinese, with an English excerpt shared here. Two young runaways follow an ancient aqueduct into the mist, through a world of moss sheep, ruined towers and dark humor. |
|
Another ongoing science-fiction novel in Chinese, with an English excerpt shared here. On a conquered ocean world, an investigator enters a dying nuclear submarine with its last surviving guide, following an anomalous sound that reaches across interstellar space. |
Data Science · Analytics Engineering · Healthcare Data · Risk Analytics · Reproducible Analytical Systems