Zurich-based startup Flexion has raised $50 million, bringing its total funding to $57.35 million. Founded in January by former Nvidia researchers, the company is developing a universal "intelligence layer" for humanoid robots that works across different robot bodies and tasks. Flexion's system uses synthetic data from physics simulations and combines linguistic processing, motion generation and full-body control based on transformer architecture—the same technology behind large language models. Unlike traditional approaches requiring manual programming or remote control, the platform enables autonomous robot operation. The company plans to licence its software annually per robot rather than building hardware itself. Fresh capital will fund a Bay Area office, team expansion in Zurich and a larger testing robot fleet. The robotics sector attracted over $10.7 billion globally by November, reflecting strong investor interest.

Ex-Nvidia engineers build office robot that actually works. * Flexion Robotics, a Swiss startup founded by former NVIDIA robotics engineers, introduced a humanoid robot designed specifically for office environments rather than factories. * AI training combines simulation, imitation learning, and reinforcement learning, allowing robots to learn reusable skills instead of being programmed for one specific task. * Demonstrated capabilities include retrieving packages, opening doors, climbing stairs, riding elevators, unpacking deliveries, filing documents, organizing workspaces, carrying supplies, and performing repetitive office tasks autonomously. * Unlike many humanoid demonstrations that rely heavily on teleoperation, Flexion claims its robots require minimal human intervention after training and can adapt learned skills to new office scenarios. * Office automation is emerging as the next commercial battleground alongside manufacturing, with companies targeting labor shortages, repetitive administrative work, and lower deployment costs for humanoids.

Flexion Robotics, a Swiss startup founded by former Nvidia researchers, has developed a system to train humanoid robots to perform complex office tasks autonomously. The company's approach combines multiple AI systems using reinforcement learning, teaching robots individual skills in simulation before deploying them in real-world settings. Unlike traditional methods relying on teleoperation, Flexion's software enables robots to execute multi-step commands independently. A demonstration shows a modified Unitree humanoid robot retrieving a parcel by navigating stairs and lifts, then unpacking items without human control. The system works by having a master AI model analyse videos of humans performing tasks, matching them with skills learned in simulation. According to CEO Nikita Rudin, the technology's key innovation is its extensive use of reinforcement learning across all software layers, from high-level planning to motor control.