Samir Agarwala

I am an alumnus of the MS CS program at Stanford University where I worked in computer vision. I currently work on machine learning for humanoid robotics at Tesla Optimus.

At Stanford, I worked with Jiajun Wu and Leonidas Guibas. I completed my undergraduate studies at the University of Michigan, where I was advised by David Fouhey. Prior to Tesla, I worked at Apple, developing machine learning algorithms for real-time gaze tracking in AR/VR systems.

Earlier in my career, I worked with Amy Cohn, Albert Berahas, and Stephen Parker at Michigan on research spanning healthcare operations, adaptive optimization, and bioinformatics.

Email  /  CV  /  Google Scholar  /  Github

I am interested in developing robust computer vision systems that can understand the world around us. Some areas of particular interest include scene understanding, learning robust visual representations and models using unstructured visual data, and neural rendering.

A novel real-world object inverse rendering benchmark to evaluate inverse rendering methods.

Specular objects such as coke cans often appear "accidentally" in images and can be used to recover scene lighting from single image observations using differentiable rendering.

Transformers are really good at integrating evidence across multiple views and producing a planar reconstruction.

Real-life variation in learning speeds of surgical trainees and decrease in available training opportunities can affect trainee competency and potentially endager patient safety.

Identifying early disease-driving events of type 2 diabetes through integrative analysis of diverse modalities.

Website based on Jon Barron's website