Website · LinkedIn · Google Scholar · Email
I work on graph-based time series forecasting, agent-based modelling, and generative deep learning.
I am a contributor to AutoGluon — Amazon’s open-source AutoML library — working on tabular models, training correctness, time-series pipelines, and packaging.
- Contributing to AutoGluon (tabular neural nets, grouped validation, time-series covariates)
- Building AMBER, a columnar high-performance ABM framework in Python
- AMBER — Polars-backed agent-based modelling; much faster than object-per-agent baselines (paper)
- LimeSoDa — soil ML benchmark datasets (Geoderma)
- DemandBench — demand time-series forecasting datasets
- CliMaPan-Lab — climate–pandemic economic modelling




