01
Dynamic prediction from irregular clinical trajectories
U-DDBART combines patient-driven biomarker measurements with interval-censored events, enabling updated individual risk estimates as new visits arrive.
View repository ↗
Open to statistical science & applied research roles
Statistical scientist · Penn State
I turn complex data into decisions you can trust.
I build Bayesian models, clinical prediction systems, and research software—connecting rigorous statistical theory to practical, reproducible tools.
Bayesian inference
Clinical prediction
Machine learning
Spatio-temporal modeling
R · Python · Stan
02 / Selected work
Methods built for real constraints.
Recent projects spanning dynamic risk prediction, mechanistic spatial models, and production-minded machine learning.
02
Scalable Bayesian agent-based modeling
A count-valued spatio-temporal model with a neural dispersal kernel and variational inference—designed to preserve mechanistic meaning while scaling estimation.
View repository ↗03
Chest X-ray classification with DenseNet-121
An end-to-end medical imaging workflow that demonstrates model training, evaluation, and reproducible implementation in Python.
View repository ↗03 / About
Rigor without the ivory tower.
I’m Xulin, a statistical scientist based at Penn State. I care about the point where a mathematically principled model meets an imperfect dataset—and still needs to produce a useful answer.
My work crosses Bayesian computation, survival analysis, spatio-temporal systems, machine learning, and reproducible research. I’m especially drawn to problems where uncertainty is not a footnote, but part of the decision.
01
Model the data-generating process
Use structure and domain knowledge before adding complexity.
02
Make uncertainty actionable
Communicate what is known, what is not, and what changes the decision.
03
Ship reproducible work
Pair methodology with readable code, validation, and clear documentation.
Working toolkit
RPythonStanBayesian computationSurvival analysisDeep learningGitReproducible research
04 / Contact
Have a consequential data problem?
I’m interested in statistical scientist, biostatistics, quantitative research, and applied scientist opportunities where careful modeling can improve real decisions.