Research Engineer – Robot Foundation Models

Role overview

Research engineering role focused on developing foundation models for general-purpose robots. You would work at the intersection of robotics research and production-oriented machine learning, connecting vision, language, robot state, action, and temporal context. The goal is to build models that generalize across tasks, environments, and robot embodiments, then verify that progress through experiments on real robotic systems.

Responsibilities

- Design, implement, and train robot foundation models and vision-language-action systems. - Develop multimodal architectures that combine visual input, language, proprioception, robot state, actions, and temporal information. - Investigate methods for manipulation, embodied intelligence, in-context learning, and rapid adaptation from demonstrations or interaction. - Explore action representations, tokenization, prediction horizons, memory, temporal modeling, diffusion and flow-based policies, autoregressive approaches, world models, and latent-action methods. - Create training strategies using imitation learning, offline learning, reinforcement learning, and large-scale pre-training across heterogeneous robot datasets. - Train models on large GPU clusters, improve efficiency and reproducibility, and build evaluations for generalization, robustness, adaptation, and long-horizon performance. - Run controlled experiments and ablations, analyze failures on physical robots, and turn findings into improved architectures, datasets, objectives, and training methods.

Requirements

- Strong ability to read current research, implement important ideas from scratch, run large experiments, and debug failures in real robotic systems. - Experience building and training difficult machine-learning or robotics systems rather than only integrating existing models. - Familiarity with multimodal learning, robot learning, embodied AI, or related areas. - Ability to evaluate whether model improvements transfer to real robot behavior, not just offline benchmarks. - A builder-oriented background as a research engineer, machine-learning engineer, robotics researcher, or research scientist; formal academic credentials are useful but demonstrated capability is emphasized.

Nice to have

- Experience with sim-to-real transfer, deploying learned policies on physical robots, distributed GPU training, or large-scale ML infrastructure. - Research publications or strong open-source contributions in robotics, machine learning, multimodal learning, or embodied AI.