AI Research Engineer – Dexterous Manipulation (Egocentric Models)

About the role

Flexion is building the intelligence layer for next‑generation humanoid robots. We are looking for an expert in dexterous manipulation and large‑scale modeling to lead the development of physical foundation models that turn egocentric video into Vision‑Language‑Action (VLA) capabilities for our robots.

Key responsibilities

• Architect and implement large‑scale egocentric video pre‑training objectives to learn hand‑object interaction representations.
• Develop multi‑modal foundation models that fuse visual perception and natural‑language instructions into robotic trajectories.
• Design generative action heads using diffusion models and flow‑matching techniques for complex human movements.
• Create algorithms that align video‑derived representations with the kinematic constraints of 20+ DoF humanoid systems.
• Apply offline reinforcement learning and high‑fidelity simulation fine‑tuning to optimise manipulation performance.
• Translate cutting‑edge research on scaling laws and world models into production‑ready robot architectures.

Required profile

• PhD or Master’s in Robotics, Machine Learning or a closely related field, focused on data‑driven manipulation, egocentric vision or foundation models.
• Hands‑on experience with humanoid or dexterous manipulation and contact‑rich physics.
• Proven track record deploying learning‑based controllers on real robot hardware.

Required skills

• Python programming
• PyTorch and distributed training (FSDP, NCCL)
• Diffusion Models, Flow Matching, Transformers
• Reinforcement Learning and simulation environments (IsaacLab, MuJoCo)

What we offer

• Competitive compensation
• Enhanced pension plan and paid‑leave perks
• Relocation and permit sponsorship
• Central Zürich office with top‑tier robotics testing facilities
• Access to Europe’s leading robotics research team and groundbreaking projects
• Energetic, collaborative culture with regular community events