Flexion Robotics

Do you want to see a humanoid AI lab from the inside? I do 🙂 – join me and let’s visit Flexion Robotics: Europe’s leading lab building the AI brains for humanoids!  Their goal? Build the Android of robotics. One operating system that works on any humanoid hardware. It is crazy what their robots can do. We pushed a robot on the stairs. It didn't care. They left one in a rainy forest it was never trained on. No issue. And because they let robots train on their own, from zero through reinforcement learning, the robots do odd things. They stand up from the ground in ways no human ever would. Because they learned what's best for a robot body, not a human one. Backflips? Surprisingly easy. Finding and picking up a box without crushing it? That's still a hard problem to enable more industry usecases. And Flexion is exactly working on that. If their humanoid brain works, this will be industry defining. This is why Flexion is one of Europe's Most Ambitious Startups. 🇪🇺🔥 Full Video on Youtube: https://lnkd.in/e-2YnrjE Subscribe to all our full updates here: https://lnkd.in/d9n3Hujn Thanks to Nikita Rudin & Julian Nubert for inviting me and showing me around.

“Teleoperation in robotics is very popular right now. We’re intentionally avoiding it.” Nikita Rudin, CEO of Flexion Robotics on how they’re training robots at scale: “We’re betting heavily on simulation and reinforcement learning.” “No motion-capture suits. No VR headsets. No armies of people piloting robots.” “Instead, we train as much as possible in simulation.” “If a new robot comes in, we load its URDF, retrain in the simulator, and deploy a new neural network.” “So when we train one robot on a task, we’re effectively training dozens or hundreds of embodiments at once.” “That’s the flywheel.”

Today, we're joined by Nikita Rudin, co-founder and CEO of Flexion Robotics to discuss the gap between current robotic capabilities and what’s required to deploy fully autonomous robots in the real world. Nikita explains how reinforcement learning and simulation have driven rapid progress in robot locomotion—and why locomotion is still far from “solved.” We dig into the sim2real gap, and how adding visual inputs introduces noise and significantly complicates sim-to-real transfer. We also explore the debate between end-to-end models and modular approaches, and why separating locomotion, planning, and semantics remains a pragmatic approach today. Nikita also introduces the concept of "real-to-sim", which uses real-world data to refine simulation parameters for higher fidelity training, discusses how reinforcement learning, imitation learning, and teleoperation data are combined to train robust policies for both quadruped and humanoid robots, and introduces Flexion's hierarchical approach that utilizes pre-trained Vision-Language Models (VLMs) for high-level task orchestration with Vision-Language-Action (VLA) models and low-level whole-body trackers. Finally, Nikita shares the behind-the-scenes in humanoid robot demos, his take on reinforcement learning in simulation versus the real world, the nuances of reward tuning, and offers practical advice for researchers and practitioners looking to get started in robotics today. 🎧 / 🎥 Listen or watch the full episode on our page: https://twimlai.com/go/760.

We’re organizing the world’s most talent-dense simulation hack in Zurich with NVIDIA and world-class Zurich startups Flexion Robotics and Jua.ai. This is the next flagship builders hackathon in Zurich. Happening from 10 October to 12 October. We will be joined by top researchers and entrepreneurs. Simulate anything. Physics, biology, digital twins, robotics, financial markets, weather/climate, psychohistory, … ✅ one H100 per participant ✅ maximal talent density ✅ top tier researchers/engineers from many fields Applications are now open. For our last robotics hack, participants flew in from Canada. The talent density will be even more insane this time. Apply here: https://luma.com/s4knila9 Organizing together with Anirudhh Ramesh.