Vikram Sundar

I am a machine learning scientist broadly interested in applications of machine learning to structural biology and drug discovery. I am currently a machine learning scientist at Generate:Biomedicines and just completed my PhD in Computational and Systems Biology at MIT, working with Professor Kevin Esvelt on machine learning for protein design and funded by the Hertz Fellowship. Specifically, I worked on a method named FLIGHTED for denoising high-throughput experimental data for protein fitness modeling and used the methods I developed alongside large quantities of experimental datapoints to engineer alternate specificities of TEV protease.

I have previously worked on a wide variety of other projects, including internship projects on diffusion models at Generate:Biomedicines and on permeability modeling with molecular dynamics at Inductive Bio. Before my PhD, I spent a year at Google Accelerated Science as an AI resident working on machine learning with DNA-encoded libraries and a year at the University of Cambridge funded by the Churchill Scholarship with Dr. Lucy Colwell working on generalization and attribution analysis of machine learning models for protein/ligand binding.