Internships
Research internships at Google, NVIDIA, Adobe and MERL.
Back to projects Google
Industry research in novel view synthesis, physical properties of 3D assets and diffusion models.
Novel view synthesis at scale
Token disentanglement lets transformer models learn from diffusion-generated views and generalize to unseen scenes.
NVIDIA
Physical-property assignment to 3D meshes
Combine foundation-model features and multimodal reasoning to segment 3D assets into meaningful parts and identify their materials.
Adobe
Control2Control
Built a training-free diffusion editing pipeline, using revitalized attention transfer to apply spatial edits while preserving the surrounding scene.
MERL
Steered Diffusion
A general framework steers pretrained diffusion models during sampling for image-to-image and label-conditioned generation.