about me.
I am a senior at Stanford studying Computer Science and Philosophy. Previously, I was a research intern at NVIDIA GEAR, advised by Jim Fan and Yuke Zhu, where we scaled world models for embodied AI and invented WAMs. Before that, I was a research intern at Together AI working on omni-model training and optimization. Feel free to reach out at ayaan04 [at] stanford [dot] edu.
Light strikes the eye, pixels fill a frame, sound arrives as pressure, but none of this is yet understanding. The world must be gathered from fragments: motion, memory, expectation, hidden causes. To truly understand is to build an inner world that abstracts reality enough to dream beyond it. My research begins where perception becomes imagination, and imagination makes action possible.
work.
DreamZero: World Action Models are Zero-shot Policies
A World Action Model built on a pretrained video diffusion backbone that jointly predicts future world states and actions, achieving over 2x improvement in generalization to new tasks and environments compared to state-of-the-art VLAs in real-robot experiments, with real-time closed-loop control at 7Hz.
DreamDojo: A Generalist Robot World Model from Large-Scale Human Videos
A foundation world model pretrained on 44k hours of diverse human egocentric videos—the largest dataset to date for world model pretraining—that demonstrates strong generalization to diverse objects and environments, with stable real-time interactions at 10 FPS for over 1 minute after distillation.