Generative AI has already transformed how we code, analyze data, and reason. The next frontier is physical: robotics. 

For decades, progress in robotics has been defined by brittle, task-specific systems that rely on scripted behaviors. Robots could work, but only in tightly controlled environments. If we want robots to truly help humanity, whether on factory floors, in disaster zones, or beyond Earth. They need to think, learn, and adapt. 

Right now, too many robots still rely on human hands behind the scenes. Thousands of hours of demonstration, hundreds of operators, millions of dollars, all to teach a single task. This doesn’t scale. That’s not progress. As large language models automate tasks involving reasoning, writing and creativity, robots should also be able to move and adapt on their own. We’re building toward that future. 

At Flexion, we’re building a reinforcement learning and sim-to-real platform that can power humanoid robots across morphologies and tasks. Our full autonomy stack spans: 

• Command Layer: language models for common-sense reasoning that take in tasks described in natural language, break them down into subtasks, and provide the necessary environment understanding and grounding. 

  
  

• Motion Layer: a vision-language-action model trained primarily on synthetic data, fine-tuned for real-world edge cases. 

  
  

• Control Layer: transformer-based, low-latency whole-body control with a modular skill library, enabling rapid composition of new behaviors. 

The result: robots that can be deployed in the real world with minimal human involvement. No scripts, no tele-op farms, no brittle task-specific logic. Just adaptive, intelligent systems that learn. 

Even with record investment and rapid progress in hardware, robotics is still at the edge of its true breakthrough. Around the world, humanoid robots look impressive, but few can perform useful, scalable work outside controlled settings. The missing piece isn’t mechanics, it’s intelligence. Flexion is building that intelligence layer, the brain that will turn capable machines into adaptable, autonomous systems.

The timing is clear. The same compute and training infrastructure that made LLMs possible is now enabling robotics to cross its own utility threshold. Meanwhile, demographic shifts and labor shortages are accelerating worldwide. A third of the developed world’s population will be over 60 by 2050, and industrial output is already feeling the impact. Humanoid robots aren’t a sci-fi fantasy, they’re an economic imperative. 

To accelerate our growth, we’re proud to share that Flexion has raised $50M in Series A funding, from DST Global Partners, NVentures (NVIDIA’s venture capital arm), redalpine, Prosus Ventures, and Moonfire. This follows $7.35M in seed funding from Frst, Moonfire, and redalpine just a few months earlier. This funding allows us to expand our Zurich R&D team, scale compute and robot fleets, establish a U.S. presence, and accelerate commercialization of our autonomy stack. We are already working with major OEM partners, and today's news will help scale these partnerships globally. 

Flexion is led by a team of world-class roboticists, engineers, and scientists with deep roots in reinforcement learning, control systems, perception, and mechatronics. Our founding team brings together alumni from ETH Zurich, NVIDIA, Meta, Google, Tesla, and Amazon, with contributions spanning cutting-edge RL research, humanoid control architectures, and large-scale simulation environments. We’ve spent our careers tackling the technical barriers that have kept robots from scaling. Flexion is the product of those hard-earned lessons. 

Our mission is simple but ambitious: to power the intelligence stack for humanoid robots, so they can work alongside humans, not depend on them. We’re not building the body. We’re building the brain. This is a hard problem. But the right kind of hard. 

If you’re a roboticist, engineer, or partner who refuses to settle for the status quo, reach out.