Uncovering temporal dynamics of large collection of cells by applying machine learning and dynamical systems framework to single cell sequencing data.
I'm a computational scientist working at the intersection of genomics, machine learning and nonlinear dynamics in advancing data-driven scientific discovery. Specifically, I'm developing a novel based approach whereby RNA velocity is used to further dissect a purely expression-based clustering using ideas of attractors and basins in dynamical systems. In another direction, I'm also working on developing an integrated mouse phenotyping platform that is an end-to-end solution to data collection, automated behavior quantification using machine learning and downstream genetic analysis for the laboratory mouse.
I earned my PhD in physics focusing on the mathematical modeling of dynamical systems (deterministic and stochastic differential equations) on graphs (complex network topologies) to study the emergence of rich spatio-temporal patterns in various complex systems spanning from physics to ecology. After PhD, I did my first postdoc in the theoretical ecology group at the Institute for chemistry and biology of marine ecosystems in Germany, where I worked on developing mathematical models of biodiversity in multi-species ecosystems competing for limited resources and developed expertise in applying machine learning methods to learn species-species trait relationships. Subsequently, I took a second postdoc at NC state university where I worked on the design and implementation of physics aware neural networks that can learn to correctly predict the dynamics of a system while respecting the underlying conservation laws. I also worked a bit on design of latent models that would enhance the interpretability and explainability of models in scientific machine learning.
At JAX, I'm excited to work on predicting cell fate dynamics by constructing a transcriptome-wide continuous vector field from RNA velocity.