Research Scientist - Robot Learning (VLA / WAM)

About the role

SpAItial is pioneering next‑generation world models that understand physics and geometry, enabling robots to act in physically‑grounded 3D environments. We are looking for a senior research scientist to own the end‑to‑end training of vision‑language‑action and world‑action models that turn these world models into competent robot policies.

Key responsibilities

• Own the full training pipeline for VLA and WAM models, from data collection to policy deployment on robots.
• Close the sim‑to‑real gap using domain randomisation, system identification and calibration, and build evaluation metrics that predict real‑world transfer.
• Adapt vision‑language model backbones for control, curate heterogeneous robot datasets, and design action representation and decoding mechanisms.
• Develop world‑model components that predict future observations conditioned on actions, and conduct post‑training fine‑tuning and reinforcement‑learning for robustness.

Required profile

• PhD in robotics, machine learning or computer vision with a focus on robot learning, or equivalent industry experience.
• Publications at top venues (CoRL, RSS, ICRA, IROS, CVPR, ICCV, ECCV, NeurIPS) and/or open‑source contributions or deployed systems.
• Proven experience designing and training modern robot policies (VLA, WAM, diffusion) end‑to‑end.

Required skills

• Strong fundamentals in imitation learning and reinforcement‑learning fine‑tuning of pretrained policies.
• Fluency with vision‑language model backbones and adapting them for control tasks.
• Expertise in Python and PyTorch, including multi‑node distributed training (FSDP or equivalent).