<aside> <img src="attachment:d1fa4233-c8ae-4d2b-890f-a46ecf5bd593:image-1783414083764.png" alt="attachment:d1fa4233-c8ae-4d2b-890f-a46ecf5bd593:image-1783414083764.png" width="40px" />
Build physical intelligence with us. We're building the foundation model for physical intelligence — teaching machines how the physical world works, not just how it looks. We're small, early, and hiring the people who will define what we become.
</aside>
<aside> ⭐
Selected for SPRIND's Next Frontier AI Challenge — one of ten teams chosen to build Europe's next generation of frontier AI labs.
</aside>
Ontic Labs is building the foundation model for physical intelligence — giving machines a working grasp of how the physical world behaves. We develop physics-aware world models that let machines perceive, predict, and act in complex real-world environments, laying the groundwork for general-purpose robotics and embodied AI.
This charts a route beyond today's data- and compute-driven paradigm: where current models capture how the world looks, ours captures how it works — learning the physics of the world once, so that every robot can perform every task in every scene.
We're at the stage where a handful of people set the direction for everything that follows, so we keep the team small and the signal high.
Ontic Labs was started by a founding team from the Max Planck Institute for Intelligent Systems, the University of Tübingen, the Tübingen AI Center, and the ELLIS Institute Tübingen — some of Europe's leading machine-learning groups — now building and scaling this foundation model together.
Ontic Labs has been selected for the Next Frontier AI Challenge run by SPRIND, Germany's Federal Agency for Disruptive Innovation — a €125M programme to build Europe's next frontier AI labs. We're one of ten teams chosen by an expert jury of leading researchers, frontier-lab veterans, and deep-tech investors, with non-dilutive funding and a staged path to as much as €26M per team.
For you, that means real runway, serious backing, and independent validation that the bet is worth making — while the team is still small enough that your work shapes everything.
We're looking for researchers with a PhD and a strong track record in learned world models — 3D scene representation, dynamics prediction, and simulation learned from video and multimodal data. In this role, you'll design and train models that capture how 3D scenes are structured and how they evolve and respond over time, build the benchmarks that measure physical forecasting, and turn geometric and temporal structure into controllable predictions. This role is for someone who moves fast from a research idea to a working model, and is excited to create the missing pieces along the way.
We're looking for a senior engineer to own the foundation the whole lab runs on — training infrastructure, large-scale data pipelines, and the systems that let a small research team move fast. In this role, you'll turn research prototypes into robust, scalable systems, make the hard build-versus-buy calls, and set engineering standards from day one. This role is for someone who treats infrastructure as a product, debugs across the whole stack, and gives researchers leverage without slowing them down.
We're looking for exceptional students — PhD, MSc, or standout undergraduates — to work directly with the founding team on core research: 3D representations, learned dynamics, video self-supervision, and robot learning. In this role, you'll own a real research question end-to-end, run experiments that feed straight into our roadmap, and ship ideas instead of watching from the sidelines. This role is for someone who is hands-on with PyTorch and large-scale training and genuinely curious about physical intelligence.
Don't see your exact role? If you're exceptional and excited about physical intelligence, tell us what you'd want to build. We hire for talent and fit over titles — surprise us.