De Wang

Principal Scientist at Amazon — multimodal foundation models for physical-world reasoning.

Sunnyvale, CA LinkedIn Google Scholar Publications

About

I'm a Principal Scientist at Amazon, where I work on multimodal foundation models that help machines perceive and reason about the physical world, to help Amazon deliver packages to customers faster and more efficiently. Before Amazon, I worked on on-device generative AI at Apple, VLM for document and webpage understanding at Salesforce Research, and perception for autonomous driving at Argo AI, and early-career research at Microsoft Research. I hold a Ph.D. from the University of Texas at Arlington.

Focus

Multimodal foundation models for physical-world reasoning — physical world understanding that generalizes beyond any one task.
World models for autonomous & robotic systems — representations that support planning and action, not just recognition.
Efficient on-device generative AI — hardware-aware attention primitives, low-bit quantization and few-step diffusion for models that run on a phone.

Path

Principal Scientist, Amazon

Multimodal Foundation Model · Current

Multimodal foundation models for physical-world, spatial temporal understanding across the last mile.

On-device Generative AI, Apple

3.5 years

Core contributor to on-device diffusion and multimodal models; shipped Image Playground and Genmoji as part of Apple Intelligence.

Document Intelligence, Salesforce Research

1 year

VLM for document and webpage understanding.

AV Perception, Argo AI

3 years

Outdoor scene understanding for autonomous driving; contributed to the perception stack, and Argoverse dataset and benchmarks.

Research Intern, Microsoft Research

Earlier

Early-career research at the intersection of machine learning and security.

Contact

Research collaborations/conversations, or ideas of build something cool in digital or physical-world AI are all welcome. The fastest way to reach me is via email.