Sharut Gupta
I am a fourth-year Ph.D. student at MIT CSAIL, advised by Prof. Phillip Isola and Prof. Stefanie Jegelka. I have spent time at Google DeepMind (Gemini) and at Meta Superintelligence Labs (FAIR), working on post-training large language models. Previously, I completed my undergraduate studies in Mathematics and Computing at the Indian Institute of Technology, Delhi (IIT Delhi), where I was mentored by Prof. Yoshua Bengio for my thesis.
My long-term research goal is to build intelligent systems that can understand, integrate, and reason over continuous, multimodal, real-world sensory inputs. To this end, I focus on two key intertwined paths:
- Representation Learning: How do we pretrain models to learn from diverse, unpaired, heterogeneous data and discover a grounded representation that enables multimodal understanding?
- Adaptive Intelligence: How do we design algorithms that enable efficient and robust adaptation under continuous distribution shifts, changing users, and novel tasks?
What's New
Check out the latest news about my research, talks and more.
-
Excited to organize the Any-to-Any Multimodal Learning (A2A-MML) Workshop at CVPR 2026!
-
Our work on Unpaired Multimodal Learning got accepted at ICLR 2026! (paper)
-
Our work on Prefix-Scannable Models got accepted at ICLR 2026! (paper)
-
Gave an invited talk at the AI, ML and Computer Vision Meetup hosted by Microsoft and Voxel51 (recording).
-
Serving on the panel for Artificial General Intelligence (AGI) at NCRC 2026 at Harvard University.
-
Excited to attend and give a talk at the Aspen Meeting on Foundation Models!
-
Our work on Representation Guidance for Diffusion Models got accepted at NeurIPS 2025! (paper)
-
Awarded the MathWorks Engineering Fellowship; thank you, MathWorks!
-
Thrilled to be interning at Meta Superintelligence Labs this summer with Mohammad Pezeshki and Mark Ibrahim!
-
Got featured in MIT's CSAIL Alliances Student Spotlight!
-
Recieved the finalist award (top 3) for the Citadel GQS PhD Fellowship.
-
Received the Citadel Securities PhD Summit Award.
-
Our work on Learning Disentangled Multimodal Representations got accepted at ICLR 2025! (paper)
-
Gave a talk at the NeurIPS 2024 Workshop on Self-Supervised Learning â Theory and Practice (recording).
-
Invited to speak at the MIT CSAIL Embodied Intelligence Seminar (recording).
-
Our NeurIPS 2024 paper, In-Context Symmetries was featured by MIT News!
-
Received the Top Reviewer at NeurIPS 2024.
-
Received the Honorable Mention Award at the UniReps Workshop, NeurIPS 2024.
-
Recognized as a finalist for the Jane Street Graduate Research Fellowship.
-
Received the MIT Presidential Fellowship!
-
Our work on In-Context Symmetries got accepted at NeurIPS 2024! (paper)
-
Our recent paper on the role of equivariance in SSL got accepted at NeurIPS 2024! (paper)
-
Gave a talk at the MIT Machine Learning Tea (ML Tea) Seminar series.
-
Invited to present at TAG-DS Pacific Northwest Seminar on Topology, Algebra, and Geometry in Data Science.
-
Spoke at the Quantitative Translational Imaging in Medicine (QTIM) Lab, Harvard University.
-
Presented at the MIT LIDS and STATS Tea Talk series.
-
Thrilled to be interning at in the Gemini Team at Google DeepMind this summer with Dilip Krishnan!
-
Our recent paper, Context is Environment got accepted at ICLR 2024! (paper)
-
Our paper on rotationally equivariant contrastive learning got accepted at ICLR 2024! (paper)
-
Our work on removing biases from molecular representations got accepted at ICLR 2024! (paper)
-
-
Gave a talk at the NeurIPS 2022 Workshop on Federated Learning: Recent Advances and New Challenges (recording).
-
I have officialy started my PhD at MIT!
