Dream

Dream is a research-and-product lab based in San Francisco working at the intersection of generative AI and the built environment. The company trains generative models over structured, physically grounded representations of how buildings and infrastructure are assembled, with the aim of producing 3D architectural and engineering layouts that are structurally coherent and buildable.

Its principal product, Dream architectural and engineering intelligence, is a system of generative models designed to understand how the physical world is assembled, how it functions, and how it can be transformed. The output is 3D layouts aligned with real-world design patterns, intended for use across the architecture, engineering and construction (AEC) sector, including architecture, engineering and construction workflows.

The work draws on technical domains including generative AI, machine-learning research, 3D layout generation, spatial reasoning, structured relational data modelling and physical world modelling. Dream describes itself as combining frontier machine-learning research with real-world product experience, and its team is multidisciplinary, spanning engineers, researchers, product leaders and architects.