We are building a general AI for humanoid robots that learns in the real world with minimal human supervision. Our approach is based on world-model-based reinforcement learning trained on large-scale data that is practical to collect. This makes it possible to learn reliable behaviors and improve beyond human performance limits. As a founding engineer, you will design and build core systems while helping shape the technical direction from the ground up.
Responsibilities
• Build data pipelines that scale to 100k+ hours of multimodal robot data
• Scale pre- and post-training runs on 1000s of GPUs
• Improve inference efficiency on low-power embedded systems
• Lead core architectural decisions, help define the long-term technical roadmap, and set engineering standards from the beginning
Qualifications
• Strong practical experience building software infrastructure
• Experience implementing and debugging distributed systems
• Extensive experience in Python and at least one deep learning library such as PyTorch or JAX
• Ideally experienced in implementing scalable training pipelines for world-model-based RL
• Experience with systems programming languages (e.g. Rust, C++) is a plus
We're a small, focused team where your contributions would have a major impact. As a founding team member, you'd also have significant ownership in what we're building together - we believe in giving our early engineers meaningful equity that reflects their foundational role.
