Machine Learning Engineer – Multimodal Foundation Models

About the role

We are building unified foundation models that reason across text, image, video, and kinematics to power intelligent robotic policies. As a Machine Learning Engineer you will own the end‑to‑end pipeline from data collection to model deployment, working on large‑scale multimodal networks that run on real robots.

Key responsibilities

• Design and implement native multimodal architectures that share a unified representation across vision, language, and other modalities.
• Research and develop cross‑modal reasoning techniques that go beyond simple association, e.g., grounding visual physics in kinematic constraints.
• Own the full training loop: set up large‑scale experiments, debug failures, improve data mixtures, and tighten evaluation metrics.
• Integrate trained models into the robot software stack, optimize for edge inference, and iterate on real‑world deployments.

Required profile

• Exceptional mental acuity and ability to learn and reason quickly across unfamiliar domains.
• Strong engineering curiosity with a track record of digging into system internals.
• High‑performance mindset: comfortable moving fast, handling ambiguity, and delivering in demanding environments.

Required skills

• Very strong coding skills in Python, C++, or Rust.
• Production experience training and deploying large‑scale multimodal models.
• Deep understanding of LLM‑style pretraining, post‑training, and reinforcement learning at scale.
• Fluency with large GPU clusters and infrastructure for massive experiments.

What we offer

• Opportunity to work with a small, elite team on cutting‑edge AI for robotics.
• Broad ownership and radical responsibility across the full stack.
• Fast‑paced environment that rewards speed, intelligence, and deep engineering instinct.