Rohit Dilip

I research generative modeling for biology at Caltech. Recently, I've worked on using flow matching to develop scalable tokenizers and generating new biological structures using autoregressive models. Before grad school, I studied quantum information at TUM and got my bachelor's degree in physics from Princeton.
I promise I am usually smiling.

Recent Publications

Kanzi: Flow Autoencoders are Effective Protein Tokenizers

Rohit Dilip, Evan Zhang, Ayush Varshney, David Van Valen
We develop a flow matching based autoencoder for protein structure tokenization. Kanzi outperforms 20x larger models trained on 400x as much data.

Jovian: Bidirectional Autoregressive Protein Structure Generation

Rohit Dilip, Ayush Varshney, Evan Zhang, David Van Valen
Jovian presents an autoregressive approach to condition generative models on particular substructures.

CellSAM: A Foundation Model for Cell Segmentation

Uriah Israel, Markus Marks, Rohit Dilip, Qilin Li, et al.
CellSAM is a generalist model for cell segmentation that works across visual modalities, sizes, and shapes.

Data compression for quantum machine learning

Rohit Dilip, Leo Li, Adam Smith, Frank Pollmann
Physical Review Research, 2023
We develop a tensor-network based data compression technique specifically for exploitation on low-depth quantum circuits.

Real- and imaginary-time evolution with compressed quantum circuits

ShengHsuan Lin, Rohit Dilip, Adam Smith, Frank Pollmann
PRX Quantum
We show that low depth classical simulations of quantum circuits can be used to simulate non-equilibrium quantum dynamics.

Research Notes

Loading posts...