Research Engineer – Robotics & Embodied AI

Description

The company is an early-stage AI lab building frontier multimodal models. It was founded by former DeepMind researchers, one of whom, Andrew Dai, previously led work on Gemini. The team has 20 scientists and engineers and is growing quickly.

The company raised $55M in seed funding from Striker Ventures, Menlo Ventures, Altimeter Capital, and NVIDIA.

As a foundation model company, the team wants to help robotics developers with visual reasoning, spatial understanding, and high-quality robotics data. The role suits people with a robotics background who want to apply that knowledge to large-scale foundation models. Rather than working only on single robot hardware or narrow policies, you will work with the core foundation model research team to bring physical-world knowledge into the models, solve practical robotics problems, and build capabilities that transfer across many robotics applications.

Responsibilities

• Improve and measure visual reasoning, 3D spatial understanding, and physical affordance inside the multimodal foundation models, to support robotic perception and control.
• Design scalable data pipelines covering collection, teleoperation, trajectory annotation, automated quality checks, and evaluation for multimodal robot data.
• Research and build new methods for robotic task planning, hierarchical reasoning, and decision-making using multimodal foundation models and Vision-Language-Action (VLA) architectures.
• Work with foundation model researchers and robotics partners to find real-world bottlenecks, assess model abilities, and build foundational AI tools for the robotics field.

Requirements

• A robotics background from research or engineering, with hands-on work in robotic perception, planning, control, or robot learning.
• Real interest in moving robotics knowledge into large-scale deep learning, multimodal architectures, and foundation model research.
• Solid machine learning experience and hands-on familiarity with modern LLMs, vision-language models (VLMs), or transformer architectures.
• Strong research and problem-solving skills, and the ability to work independently on open-ended, complex technical problems.
• Experience with robotics simulators such as Isaac Sim or MuJoCo, for generating large-scale data, evaluating policies, or pre-training models.
• Experience with robotics datasets, trajectory annotation, task decomposition, and benchmark evaluation for physical systems.
• Experience applying, fine-tuning, or building LLMs, VLMs, or Vision-Language-Action (VLA) models for robotics.
• Hands-on work with real robots in manipulation, navigation, or autonomous agents, and an understanding of the practical limits of running policies on physical hardware.
• Deep expertise in robotic planning, 3D spatial reasoning, affordance modeling, or embodied intelligence.