We are a computational lab with interests in machine learning, mechanistic modeling, and AI for Science. Our goal is to understand how structure and dynamics give rise to function in complex systems – from molecules and cells to circuits and behavior.
We develop methods that capture the shape, symmetry, and flow of biological processes in high-dimensional, often noisy data. To do this, we design algorithms in geometric deep learning, topological data analysis, and manifold learning, that relate the ‘shape’ of data to its underlying biology. We tackle challenging inverse problems – using data to infer the cellular and molecular interactions or neural connectivity that drive observable behavior. Ultimately, our aim is to create biologically informed, interpretable, AI models that not only explain how systems work but also drive translational impact through engineering and precision medicine.
- Identifying in-silico biomarkers of neurological and psychiatric conditions using multimodal neural and behavioral data
- Integrating and translating across brain imaging modalities
- Learning high-order interactions and network dynamics from large-scale brain activity recordings
- Building brain encoding and decoding models to understand how information is represented, transformed, and recalled
- Designing new learning algorithms that emulate neuronal dynamics, plasticity, and self-organization for more biologically grounded AI
