About - Nunchux AI | Nunchux AI

Where the Future of AI Media Runs

The fastest infrastructure for image, video, and world models.

We build full-stack infrastructure that pushes generative media past the limits of today's hardware.

We help developers and enterprises run image, video, and world models up to 100× faster and cheaper, at the same fidelity.

Nunchux AI is building the next frontier of multimodal inference. Founded by veterans from AMD, NVIDIA, and Snowflake, alongside world-class researchers from Berkeley, CMU, and MIT, we are inventing the infrastructure layer for the next generation of generative AI.

Founders and Advisors

Muyang Li

Expert in efficient generative models since 2020. Developed SVDQuant, Radial Attention, and Sparse VideoGen.

Yujun Lin

CTOMIT PhD, previously NVIDIA

Former NVIDIA researcher; efficient ML systems since 2018. Developed SVDQuant, QServe, HAQ, and Deep Gradient Compression.

Zhekai Zhang

Chief System ArchitectMIT PhD

Efficient ML systems since 2019. Developed SVDQuant, QServe, and Once-for-All.

Jun-Yan Zhu

Chief ScientistAssociate Professor at CMU

Pioneer of image and video generation. Previously UC Berkeley and MIT.

Alexei A. Efros

Alexei A. Efros

AdvisorProfessor at UC Berkeley

Pioneer of data-driven computer vision and graphics.

Built by Top AI Researchers and Engineers from:

Be at the heart of where creativity meets rigor.