I am a Ph.D. student at Stanford University, advised by Kayvon Fatahalian. My interests include LLM self-improvement and controllable diffusion models. Prior to my Ph.D, I received a B.S. in Computer Science at Stanford. In the past, I have worked with Pat Hanrahan on metropolis light transport and path guiding, with Stefano Ermon on applications of GANs to satellite images, and with Susan Athey on the dynamics of Bitcoin usage.

Check out my Substack for thoughts on self-improving LLM agents. New posts (bi)weekly.

Learning Agents

Self-Generated In-Context Examples Improve LLM Agents for Sequential Decision-Making Tasks
Vishnu Sarukkai, Zhiqiang Xie, and Kayvon Fatahalian. Arxiv, 2025.
^{[Paper] [Blog] [Code (coming soon)]}

Automated Rewards via LLM-Generated Progress Functions
Vishnu Sarukkai, Brennan Shacklett, Zander Majercik, Kush Bhatia, Christopher Ré, Kayvon Fatahalian Arxiv, 2024.
^{[Paper] [Blog] [Code (coming soon)]}

MLMove: Learning to Move Like Professional Counter-Strike Players
David Durst, Feng Xie, Vishnu Sarukkai, Brennan Shacklett, Iuri Frosio, Chen Tessler, Joohwan Kim, Carly Taylor, Gilbert Bernstein, Sanjiban Choudhury, Pat Hanrahan, Kayvon Fatahalian. SCA, 2024.
^{[Paper] [Code] [Project Page]}

Controllable Image Generation

Block and Detail: Scaffolding Sketch-to-Image Generation
Vishnu Sarukkai, Lu Yuan, Mia Tang, Maneesh Agrawala, Kayvon Fatahalian. UIST, 2024.
^{[Paper] [Code] [Project Page] [Demo]}

Collage Diffusion
Vishnu Sarukkai, Linden Li, Arden Ma, Chris Ré, and Kayvon Fatahalian. WACV, 2024.
^{[Paper] [Code] [Blog]}

Sketch-a-Sketch: Controlling diffusion-based image generation with just a few strokes
Vishnu Sarukkai, Chris Ré, and Kayvon Fatahalian
^[Blog]

Efficient Model Validation

Low-Shot Validation: Active Importance Sampling for Estimating Classifier Performance on Rare Categories
Fait Poms*, Vishnu Sarukkai*, Ravi Teja Mullapudi, Nimit S. Sohoni, William R. Mark, Deva Ramanan, and Kayvon Fatahalian. ICCV, 2021, Oral Presentation.
^{[Paper] [Blog]}

Forager: Rapid Data Exploration and Model Development
Fait Poms, Mihir Garimella, Vishnu Sarukkai, Ravi Teja Mullapudi, Jeremy Barenholtz, and Kayvon Fatahalian
^[Demo]

Miscellaneous

Cloud Removal from Satellite Images using Spatiotemporal Generator Networks
Vishnu Sarukkai*, Anirudh Jain*, Burak Uzkent, and Stefano Ermon. WACV, 2020.
^{[Paper] [Code] [Dataset]}

Bitcoin Pricing, Adoption, and Usage: Theory and Evidence
Susan Athey, Ivo Parashkevov, Vishnu Sarukkai, and Jing Xia. Arxiv, 2016.
^[Paper]