Machine Learning Engineer, Flood Risk
Saltmarsh models flood risk at building-level resolution across Europe. You would work with the hydrologists on the modelling pipeline — feature engineering from terrain and rainfall data, training, and the validation harness that decides whether a model ships.
The hard part is not the model. It is that the ground truth is sparse, biased toward places that already flooded, and the consequences of being confidently wrong are somebody's insurance premium.
Python, PyTorch, and a great deal of geospatial data.
About Saltmarsh
Climate and flood-risk data for insurers and planners. Half the team are scientists, half are engineers, and the interesting work happens where those two argue.