AI Research Scientist - United States

About Flexion:

At Flexion, we’re building the intelligence layer powering the next generation of humanoid robots. Our mission is to accelerate the transition from fragile prototypes to real-world humanoid deployment. We are founded by leading scientists in robot reinforcement learning (ex-Nvidia, ex-ETH Zürich), and backed by leading international VC firms. In just months, we’ve gone from our first line of code to deploying real humanoid capabilities.

Fifty of the world's best robotics researchers are already building the future at Flexion's headquarters in Zurich. Among them are Nikita Rudin, David Hoeller, Julian Nubert, and Korrawe Karunratanakul - scientists who have redefined what humanoid robots can do. Now we are building their counterpart team in the United States, in San Francisco.

The Role:

If you are among the most capable, and ambitious robotics researchers in the country, someone who has stared at the limits of what robots can do today and decided that simply wasn't good enough, we want to hear from you.

You will join the first wave of US research scientists at Flexion, set the agenda together, own the problems that matter most, and deploy real solutions on real hardware. You will work directly with a world class team across two continents on challenges that have no published answers yet.

Key Responsibilities:

• Drive original research in learning-based robotics - from concept to hardware deployment
• Own your research direction: define problems, design experiments, and iterate fast
• Collaborate directly with Flexion’s Zurich research team across Manipulation, Motion Generation, and Perception
• Contribute to a shared culture of scientific rigour, velocity, and impact

Requirements

• PhD in Robotics, Machine Learning, or a closely related field - completed or near completion
• Proven track record of deploying learning-based controllers on real robotic hardware
• Expert-level knowledge in:
• Reinforcement learning
• Physics-based simulation (Isaac Gym/Lab, MuJoCo, or equivalent)
• Python and PyTorch, including training large-scale neural networks
• Deep expertise in at least two areas of the following:
• Diffusion models
• Flow matching
• Dexterous manipulation
• Sim-to-real transfer and real-to-sim calibration
• Whole-body control and loco-manipulation
• Synthetic data generation for robot learning
• Vision Language Model Fine-Tuning (SFT and RL-based)
• Transformer-based 3D Scene Understanding

Benefits

• Competitive compensation package
• Comprehensive health and retirement plan benefits
• A front-row seat at one of the world’s most ambitious robotics companies
• An energetic, collaborative team with a relentless bias for action
• The opportunity to build something no one has ever done in this field - alongside the world’s leading researchers