I'm a PhD student at Mila and University of Montreal, advised by Aishwarya Agrawal. This summer I'll join Qualcomm Research in Amsterdam as an intern. I work on generative models, world models, and reinforcement learning. My research is supported by the Fonds de recherche du Québec – Nature et technologies (FRQNT).
Research
Scaling on internet text has taken us remarkably far, and the path toward physical intelligence is now becoming clearer. I am inspired by the Era of Experience vision for scalable intelligence and excited about research in physical AI, particularly improved training and representation learning in generative models. To build agents that can operate effectively in large, complex environments, we need better representations and more scalable policy-gradient algorithms for learning from experience.
A few deep-learning "topics" I like the most: diffusion and flow-based models; the target network trick from RL; and SSL, including asymmetric views in BYOL.
Selected publications
See Google Scholar for the full list. *denotes equal contribution.
-
One Flow-Transformer for Imagination and Control
-
Grounding Computer-Use Agents from Demonstrations
-
The Promise of RL for Autoregressive Image Editing
-
Rendering-Aware RL for Vector Graphics Generation
-
CTRL-O: Language-Controllable Object-Centric Representations
-
VisMin: Visual Minimal-Change Understanding
-
Hard Negatives to Enhance Visio-Linguistic Compositional Understanding
Misc
Talks I put extra care into, and a bit of teaching/organizing on the side.
-
Few-step diffusion modeling
-
Score-based generative models and diffusion models
-
-
IFT 6765 – Links between Computer Vision and Language