Lumen Labs is building the cognitive layer for physical AI. Nearly every major robotics effort today teaches machines by showing them thousands of hours of human teleoperation. We think that paradigm has a ceiling, and that the path to robots that genuinely adapt runs through a different kind of learning. Read our white paper here.
We are a small yet scrappy team of two (you’ll be teammate #3) that is quirky, humble, and impassioned. We’ve been founders in the past and we believe we are doing our most important work yet. We’re pre-seed with runway well into 2027, raising our seed later this year.
Location: Remote to start, then in-person in San Francisco from fall 2026. We’re happy to figure out visa and relocation for the right candidate.
We are currently training and testing a lot of different models against a lot of different sims and testing benchmarks. Hooking up models to a sim has been a huge pain point for us. We use Gymnasium and Shimmy as the plumbing that makes heterogeneous sims and models talk to each other at high efficiency (if you think there's a better way we're all ears). As we build, we plan to upstream useful pieces back to the Farama Foundation projects.
<aside> 🔥
You'll thrive here if you've built and run simulation environments for robotics, and have done off-the-beaten path side projects along the way.
</aside>