About Me

I'm a third-year PhD student at MIT, advised by Phillip Isola. I am supported by the NSF Graduate Research Fellowship, and previously by the MIT HDTV Grand Alliance Fellowship. I am broadly interested in generative models and representation learning, particularly data-centric approches. At the moment I'm working on the following questions:

1. Learning from synthetic data: How can we harness generative models as a data source for representation learning?
2. Human-aligned representations: How can we better align representations with human perception and perceptual preferences?
3. Science of training data: How can we better understand, improve, and sample from training datasets to enable more efficient learning and scaling?

Previously I got my bachelors in Computer Science and in Mathematics at MIT, working at the MIT Center for Brains, Minds, and Machines with Pawan Sinha and Xavier Boix. In the past I've also been fortunate to intern at DeepMind (large language models), D. E. Shaw (reinforcement learning research), Apple (applied machine learning), and Two Sigma (software engineering). In my free time I enjoy hiking, running, and tennis.

Recent News

• [Apr 2025] Gave a talk at the Stanford NeuroAILab.
• [Mar 2025] Gave a talk at Cohere for AI (recording here).
• [Jan 2025] Our paper Personalized Representation from Personalized Generation was accepted to ICLR 2025.
• [Dec 2024] Our preprint What Makes for a Good Stereoscopic Image is on Arxiv.
• [Dec 2024] Attending NeurIPS! I'm presenting our work on perceptual alignment at the main conference.
• [Oct 2024] Gave a talk at the ECCV "Efficient Text-to-Image and Text-to-3D Modeling Tutorial" (also co-organized) on Evaluating Text-to-Image Models.
• [Sep 2024] Our paper When Does Perceptual Alignment Benefit Vision Representations? was accepted to NeurIPS 2024.
• More...
[Jul 2024] Attending the International Computer Vision Summer School (ICVSS) 2024.
[Jun 2024] Attending CVPR! Co-organizing the Synthetic Data for Computer Vision Workshop
[Dec 2023] Attending NeurIPS! I'm presenting DreamSim at the main conference.
[Oct 2023] Gave a talk on DreamSim at Adobe.
[Sep 2023] Our paper DreamSim: Learning New Dimensions of Human Visual Similarity using Synthetic Data was accepted to NeurIPS 2023.
[Jul 2023] Gave a talk on DreamSim at Voxel51.
[Sep 2022] Started my PhD at MIT!
[Apr 2022] Awarded the NSF GFRP and HDTV Grand Alliance Fellowships.

Publications

* indicates equal contribution

	What Makes for a Good Stereoscopic Image? Netanel Y. Tamir*, Shir Amir*, Ranel Itzhaky, Noam Atia, Shobhita Sundaram, Stephanie Fu, Ron Sokolovsky, Phillip Isola, Tali Dekel, Richard Zhang, Miriam Farber CVPR Computer Vision for Metaverse Workshop, 2025. Paper
	Personalized Representation from Personalized Generation Shobhita Sundaram*, Julia Chae*, Yonglong Tian, Sara Beery§, Phillip Isola§. ICLR, 2025. Paper Website Code Data
	When Does Perceptual Alignment Benefit Vision Representations? Shobhita Sundaram*, Stephanie Fu*, Lukas Muttenthaler, Netanel Tamir, Lucy Chai, Simon Kornblith, Trevor Darrell, Phillip Isola. NeurIPS, 2024. Paper Website Code
	DreamSim: Learning New Dimensions of Human Visual Similarity using Synthetic Data Stephanie Fu*, Netanel Tamir*, Shobhita Sundaram*, Lucy Chai, Richard Zhang, Tali Dekel, Phillip Isola. NeurIPS, 2023 (spotlight). Paper Website Code
	Recurrent connections facilitate symmetry perception in deep networks Shobhita Sundaram*, Darius Sinha*, Matthew Groth, Tomotake Sasaki, Xavier Boix Scientific Reports, 2022. Workshop on Generalization Beyond the Training Distribution in Brains and Machines, ICLR 2021. Paper Code Poster
	GAN-based Data Augmentation for Chest X-ray Classification Shobhita Sundaram*, Neha Hulkund* Workshop on Applied Data Science for Healthcare, KDD 2021 Paper
	Do Neural Networks for Segmentation Understand Insideness? Kimberly Villalobos*, Vilim Stih*, Amineh Ahmadinejad*, Shobhita Sundaram, Jamell Dozier, Andrew Francl, Frederico Azevedo, Tomotake Sasaki, Xavier Boix Neural Computation, 2021 Paper Code

Invited Talks

• Stanford NeuroAILab, April 2025. Representation Learning with Perceptual Alignment.
• Cohere for AI, March 2024. Personalized Representation from Personalized Generation.
[Recording]
• Efficient Text-to-Image and Text-to-3D modeling Tutorial, ECCV 2024. Evaluating Text-to-Image Models.
[Slides]
• Adobe GenTech Seminar, October 2023. DreamSim: Learning New Dimensions of Human Visual Similarity using Synthetic Data.
• Voxel51 Computer Vision Meetup, July 2023. DreamSim: Learning New Dimensions of Human Visual Similarity using Synthetic Data.

Service

• Organizer: Efficient Text-to-Image and Text-to-3D modeling Tutorial, ECCV 2024
• Organizer: Synthetic Data for Computer Vision Workshop, CVPR 2024
• Reviewer: ECCV 2024