Adrian Aoun (left) and Sebastian Thrun (right). Computer scientist Sebastian Thrun, who led multiple AI projects at Google, has co-founded a brand-new startup focused on developing AI models for the engineering field, according to people familiar with the matter.
Thrun and co-founder Adrian Aoun established Dulo in San Francisco earlier this year, and the company recently closed a $130 million seed funding round, the people said. Investors in the round include former Google CEO Eric Schmidt, Kleiner Perkins chairman John Doerr, as well as venture firms DST Global, Construct Capital, and Lowercarbon Capital.
Thrun is the founder of Google X, the moonshot lab, a co-founder of Google Brain, one of Google's early AI projects, and the initiator of the self-driving project that later became Waymo. Aoun has invested in hundreds of startups, with Dulo being one of them; before founding Dulo, he had already started several companies, including the AI healthcare firm Torch (which OpenAI plans to acquire for about $100 million). Aoun also founded the robotics company Foundational (developing automated factories) and Seneca (developing firefighting drones). Former Autodesk CEO Carl Bass serves as an advisor to Dulo.
Dulo is developing AI models to automate engineering design and verification processes. For example, a user can input relevant parameters and instruct the model to design an airplane or a car, and the model will output a feasible design; the actual manufacturing is left to other companies. The startup's website describes itself as developing a "foundational large model for the hardware design field."
Dulo already faces competition from other startups in the engineering automation space. For instance, Jeff Bezos-backed Prometheus is developing AI for engineering and manufacturing; Larry Page's new company Dynatomics is building large language models for hardware engineering. Models from leading AI companies such as OpenAI and Anthropic are also continuously iterating their mechanical engineering capabilities. For example, after OpenAI launched GPT-6 Astra, users quickly discovered that the model excels at working with engineering software to model mechanical components: it can generate a computer-aided design (CAD) model of a traffic light based on an image. Even so, general-purpose models like Astra are still far from being able to design an entirely new vehicle from scratch. This also means that new entrants like Dulo have an opportunity to focus on honing models' specialized engineering capabilities and catch up.
Comments