Seahorse: A Unified Benchmarking Framework for Spatiotemporal Event Modeling
A shared framework for training and evaluating spatiotemporal event models under a consistent, reproducible protocol.
About me
I’m a machine learning researcher at DFKI with a background in applied mathematics, working on how to model events in space and time and how to test those models fairly, through open benchmarks like Seahorse.
Selected publications
Details & citationsChaotic and complex dynamics expose the limits of counterfactual reasoning
How noise, parameter uncertainty, and chaos affect counterfactual estimates in dynamical systems.
When Context Compensates for Sparse Event History: AlphaEarth for Spatio-Temporal Point-Process Forecasting
Studying geographical context for emergency medical service forecasting when local event histories are limited.
HawkesNest: A Multi-Axis Synthetic Benchmark for Spatiotemporal Pattern Complexity
Controlled synthetic event patterns for testing spatiotemporal models against known generative structure.
Get in touch
I welcome research collaborations across academia and industry. Students interested in event modeling, point processes or counterfactual reasoning are welcome to get in touch.