About the job
Flexion Robotics develops intelligence for next-generation humanoid robots, aiming to move quickly from prototypes to real-world use. Founded by specialists in robot reinforcement learning from organizations like Nvidia and ETH Zürich, the company has progressed from early code to functional humanoid systems, supported by international venture capital.
Role overview
This AI Research Engineer position centers on reinforcement learning for manipulation in humanoid robots. The work blends research and engineering, with a focus on enabling robots to perform long-horizon, contact-rich tasks reliably outside the lab. The role involves significant ownership and close collaboration with simulation and control teams to deliver integrated robotic capabilities.
What you will do
- Advance RL-based manipulation: Design and implement learning strategies for complex, dexterous manipulation tasks, moving humanoid robots toward practical deployment.
- Address long-horizon and sparse-reward challenges: Develop algorithms and systems to help robots handle multi-stage tasks with delayed or sparse rewards, tackling issues like credit assignment, exploration, and stability in real-world conditions.
- Facilitate sim-to-real transfer: Create methods for transferring policies from simulation to physical robots, focusing on robustness, managing model mismatches, and applying techniques such as domain randomization and system identification.
- Enhance RL systems: Optimize large-scale training pipelines for learning across many parallel environments, and work with simulation teams to improve efficiency and enable more complex behaviors.
- Innovate learning methodologies: Explore and apply new approaches that combine reinforcement learning, imitation learning, and model-based methods, going beyond standard practices when needed.
Location
This role is based in Zürich, Zurich, Switzerland.
