AI Research Engineer - Dexterous Manipulation (Egocentric Models)

Flexion Robotics in Zürich seeks a Senior AI Research Engineer to lead the development of foundation models for humanoid robots using egocentric video data. The role focuses on creating Vision-Language-Action models for dexterous manipulation in unstructured environments.

This role involves leading the creation of physical foundation models for humanoid robots, leveraging internet-scale egocentric video data to build Vision-Language-Action systems. The candidate will work onsite in Zürich to develop and deploy advanced learning-based controllers, taking ownership of model architecture and integrating egocentric priors with real-time robot policies for reliable manipulation.

Responsibilities

• Design and implement large-scale pre-training methods for egocentric video data to capture hand-object interactions and spatial-temporal dynamics.
• Create and scale multi-modal foundation models that combine visual input and language instructions into executable robot movements.
• Build and optimize generative action components using Diffusion Models and Flow-matching to represent complex human motion patterns.
• Develop algorithms to align human-centric video data with the kinematic limits of 20+ DoF humanoid robots for smooth, stable execution.
• Apply Offline Reinforcement Learning and simulation-based fine-tuning to improve model performance for reliable physical manipulation.
• Adapt recent research on scaling laws and world models into practical architectures that boost robot autonomy and dependability.

Requirements

• PhD or Master’s degree in Robotics, Machine Learning, or a related field, with emphasis on data-driven manipulation, egocentric vision, or foundation models.
• Experience in humanoid or dexterous manipulation, including knowledge of contact-rich physical interactions.
• Strong proficiency in Python, PyTorch, and distributed training of large neural networks (FSDP, NCCL).
• Demonstrated expertise in Diffusion Models, Flow Matching, and Transformer architectures.
• Practical experience deploying learning-based control systems on real robot hardware.
• Experience with Reinforcement Learning and simulation tools such as IsaacLab or MuJoco.

What the company offers

• Competitive compensation package
• A front-row seat at one of Europe’s most ambitious robotics companies
• An energetic, collaborative team with a bias for action

About the company

Flexion Robotics is developing the intelligence layer for next-generation humanoid robots, aiming to move quickly from prototypes to real-world deployment. Founded by experts in robot reinforcement learning (former Nvidia and ETH Zürich researchers) and backed by top international venture capital firms, the company has rapidly progressed from initial code to real humanoid capabilities in just months.

• Focus on accelerating humanoid robot deployment from prototypes to real-world applications
• Founded by leading scientists in robot reinforcement learning
• Backed by top international VC firms
• Rapid progress from first code to real humanoid capabilities in a short timeframe