Attended the Genesis Mission hands-on hackathon at Argonne National Laboratory.
Had a great week at the Junior Scientist Workshop on Theoretical Biophysics at Janelia Research Campus, getting to know the theoretical biophysics community. I gave a talk and a tutorial, Predictive information and low-dimensional structure in dynamical systems.
New preprint: Controlling turbulent flows in compressible active nematics, with Dimitrios Krommydas, Aparna Baskaran and Cristina Marchetti. Real active suspensions patterned with light develop large density variations, so we model them as effectively compressible fluids. Compressibility then becomes a control knob for turbulence: with patterned activity, it steers turbulent flow from high- to low-activity regions, and at a sharp activity boundary it holds the flow in a single chain of vortices.
Back at my alma mater, IISc Bengaluru, to give the GranSLAM Seminar: From structure formation in active–passive mixtures to effective model discovery in physics. Thanks to Prerna Sharma for the invitation.
New preprint: Contrast encodes inductive bias, with Ilya Nemenman. Contrastive methods that learn dynamics can latch onto slowly drifting noise instead of the dynamics; we fix this with a contrastive objective within trajectories, an inductive bias that leads the model to learn the physically relevant dynamical variables and disregard the noise.
New paper: Mutual information and task-relevant latent dimensionality, with Eslam Abdelaleem, Audrey Sederberg and Ilya Nemenman, at the GRaM workshop at ICLR 2026. How many variables does a prediction actually need? We turn neural mutual-information estimators into a way to read that number off from data, with a new critic that doesn’t inflate the count.
New preprint: Bicontinuity in active phase separation, with Liang Zhao, Zvonimir Dogic, Cristina Marchetti and others. In three dimensions, active flows keep active and passive fluids interpenetrating in a bicontinuous, sponge-like structure made of active sheets, which keeps reconfiguring and never coarsens.
Invited talk at the Soft Matter Seminar, Georgia Tech: From Chaos to Control: Structure Formation in Active–Passive Mixtures. Thanks to Itamar Kolvin for the invitation.
Moved to Atlanta to start as a Tarbutton Postdoctoral Fellow in Physics at Emory University, working with Ilya Nemenman. The goal: use interpretable machine learning to learn models of biophysical and nonequilibrium soft-matter systems directly from data.
Defended my PhD in physics at UC Santa Barbara! My dissertation, Phase Separation and Interfaces in Active Liquid Crystals, is on how active fluids phase separate and what happens at their interfaces. Huge thanks to Cristina Marchetti for being an amazing advisor, to the whole Marchetti group, and to my defense committee.
