Research + Modeling

Job description

Job responsibilities

• Design and run large-scale training pipelines for multimodal / VLA systems.
• Own the full loop: data → training → evaluation → deployment on real robots.
• Develop scalable infrastructure for data ingestion, training, and iteration.
• Translate model outputs into reliable, high-performance robotic actions.
• Work hands-on with robots to debug, iterate, and improve behavior.
• Define data strategy (quality, scale, diversity) and evaluation frameworks.
• Continuously improve performance across real-world tasks and environments.

Job requirements

• Built and trained large-scale models (LLMs, VLMs, or robotics foundation models).
• Deep understanding of modern ML, training dynamics, and optimization at scale.
• Experience with distributed systems and data pipelines for large-scale training.
• Comfortable operating end-to-end: from data → model → real-world deployment (including robots).
• Domain strength in at least one: robotics, VLA systems, or training LLMs/VLMs from scratch.
• Strong Python proficiency.