Your position
You will develop generative models on single-cell and spatial genomics data, from sequencing and imaging, and apply them to human developmental and disease biology. Possible research directions include:
●
Generative models of cell–cell communication. Move beyond descriptive ligand–receptor analysis to models that predict and help understand how cells react to signaling cues.
●
Joint models of regulation and dynamics. Flow- and diffusion-based models of cellular dynamics are expressive enough to map any source to any target state, but they fall short of learning the underlying regulatory principles. We work on regularization, training on time-resolved perturbation and lineage-tracing data, and multi-modal readouts to constrain these models toward biological mechanisms.
●
Intervention design and continual learning. Beyond predicting perturbation outcomes: designing interventions that steer cells toward a desired molecular state, and updating trained models as new data arrive, toward a lab-in-the-loop setup.
These directions build on our earlier work on cellular dynamics, regulation and data integration: scVelo, CellRank, CellRank 2, moslin, moscot and RegVelo. In collaboration with partners at the hospitals, universities and institutes in Basel, we generate our own data, on the newest spatial and single-cell platforms.
Your profile
Required:
● MSc in computer science, mathematics, physics, engineering, computational biology or a related quantitative field
● Hands-on experience with biological data
● Strong Python skills and experience with PyTorch or JAX, together with the statistical grounding to reason about the models you use
● A GitHub account with code you have written
● Fluent English
Desirable:
● Experience with single-cell or spatial genomics data
● Contributions to open-source scientific software