AI Research Scientist

Role Overview

The WoRV Research team studies Robotics Foundation Models that operate universally across diverse robot embodiments. Beyond individual tasks like manipulation, locomotion, and navigation, we pursue Generalist Robot Policies — a single model that handles diverse robots and tasks — and conduct frontier research spanning VLA architectures, World Models, and large-scale robot data training.

This position is also available as a Research Fellowship (3–6 months; current PhD students welcome).

Responsibilities

• Design architectures for building Robotics Foundation Models and validate cross-embodiment generalization through large-scale training.

• Research the integration of World Model-based prediction and planning capabilities into robot policies.

• Define and solve core foundation model research problems: data scaling, representation learning, action tokenization, and more.

• Design robotics dataset scaling and build training evaluation pipelines.

• Validate the generalization performance and robustness of learned policies on real robot platforms.

• Publish research results at top venues (NeurIPS, ICLR, CVPR, CoRL, ICRA).

Qualifications

• Master's degree or above in Robotics, Computer Vision, or Machine Learning, or equivalent research experience

• First-author publications at major ML / Robotics venues (NeurIPS, ICLR, CVPR, CoRL, ICRA, etc.)

• Deep understanding of and practical experience training and fine-tuning foundation models (VLA, LLM, VLM, WAM, etc.)

• Ability to independently define research problems, design experiments, and drive conclusions

• Experience with large-scale model training in PyTorch (multi-GPU / multi-node)

Preferred Qualifications

• Research experience in Robotics Foundation Models, Generalist Policies, or Cross-embodiment Transfer

• Experience designing and operating large-scale pre-training pipelines

• Research experience in World Models, Video Generation / Prediction, or Representation Learning

• Research experience in Imitation Learning, Reinforcement Learning, or Sim-to-Real Transfer

• Open-source research project contributions or research work that gained community recognition

• Research experience scaling robotics data (e.g., UMI, Ego4D)

• Experience deploying and validating policies on real robot hardware

Who We're Looking For

• Someone who first asks "why hasn't this problem been solved yet?"

• Someone who understands both the value of a paper and the value of a product

• Someone who wants to do research that aims to be the best in the world

• Someone who solves problems creatively within the constraints of hardware