Flexion Robotics develops the autonomy stack for humanoid and human-capable robots, focusing exclusively on the software intelligence layer rather than hardware. Founded by scientists, roboticists, and engineers with backgrounds at NVIDIA, the company builds AI models through simulation and reinforcement learning to enable robots to perform everyday human tasks across diverse environments - from office spaces to space exploration missions. The company recently closed a $50 million Series A funding round to expand its research and development capabilities.
The technical approach centers on sim-to-real transfer: training AI models in virtual environments using reinforcement learning, then deploying them to physical robots with minimal human intervention. The autonomy stack is designed to be hardware-agnostic, working across different robot platforms while covering the full spectrum from command-level autonomy to low-level control. This spans both manipulation tasks and locomotion systems, positioning the software as a complete "brain" that interfaces with various humanoid robot bodies.
The company's core technical domains include robot autonomy, simulation, reinforcement learning, control systems, and sim-to-real transfer. Co-led by Nikita Rudin and David Hoeller and headquartered in Switzerland, Flexion takes a research-focused approach that emphasizes building autonomous systems requiring minimal ongoing human oversight once deployed. The software-first philosophy reflects a bet that robot intelligence - not mechanical design - is the primary constraint preventing humanoid robots from operating effectively in unstructured real-world environments.