Max Robotics — Global Service Robot Directory + US Certification

About

Yilun Du is a researcher at the intersection of generative modeling, neural scene representations, and embodied AI, whose work has significantly advanced how machines perceive, model, and interact with the physical world. He is perhaps best known for his foundational contributions to energy-based models (EBMs), with his papers on implicit generation and modeling accumulating nearly 300 citations combined, demonstrating scalable MCMC-based training techniques that unlocked EBMs for high-dimensional data. His Neural Radiance Flow (NeRFlow) framework, cited over 200 times, pioneered 4D spatial-temporal scene representation from RGB images, enabling dynamic view synthesis and video processing through neural implicit representations. Du has also made substantial contributions to robotic manipulation, introducing Neural Descriptor Fields (NDFs)—SE(3)-equivariant object representations that allow robots to generalize manipulation skills across object categories, garnering over 130 citations. His more recent work spans vision-language-action models, generalist robot learning from internet video, and integration of 3D world models into robotic systems, including contributions to the Gemini Robotics initiative. Across these domains, Du's research consistently bridges theoretical machine learning with real-world physical intelligence applications.