Former Google DeepMind researcher Andrew Dai has publicly unveiled Elorian, a startup focused on improving how AI systems understand visual prompts and real-world imagery.
Dai argues that top models still perform poorly on visual reasoning and Elorian aims to close that gap, with applications spanning architecture, robotics, and automotive systems.
The launch signals growing awareness that the next frontier of AI capability lies in reliable visual and physical world understanding.
Elorian has raised $42 million in seed funding from Sequoia Capital and General Catalyst. The company employs 18 researchers, mostly from DeepMind, Meta AI, and the Allen Institute for AI.
The technology focuses on 'spatial reasoning' — the ability to understand three-dimensional relationships from two-dimensional images. Current AI models can identify objects in photos but struggle to understand spatial relationships, perspective, and physical plausibility.