Sahil Goyal
My research focuses on efficient generative models and adaptive multimodal systems that use computation more intelligently. I recently completed an AI Residency at Google DeepMind, where I worked across Veo and Gemini.
Highlights: Google DeepMind AI Residency (2024–2026) · ICML, ECCV, WACV · ICLR, CVPR, ICCV and ACM MM workshops · ACM MM Asia Best Demo Paper · 2 US patent applications
Selected research
All publications →
Recent work in efficient visual generation.
Elastic Looped Transformers for Visual Generation
Adaptive looped architectures that make visual generation more computationally elastic.
Read paper ↗
Masked Generative Nested Transformers
Nested generative transformers designed to benefit from additional computation at decode time.
Read paper ↗
Design-o-meter
A framework for evaluating and iteratively refining the aesthetic quality of graphic designs.
View project ↗