Sahil Goyal

Making generative models smaller, faster and better

I'm a Pre-Doctoral Researcher / AI Resident at Google DeepMind, where I work on making generative models like Veo and Gemini efficient. My recent work on adaptive inference for visual generation was accepted at ICML 2025. Previously, at Adobe Research, I built Design-o-meter (WACV 2025) — a first-of-its-kind framework that evaluates and refines graphic designs. My earlier research on emotionally enhanced talking faces won the Best Demo Paper at ACM MM Asia 2022.