Machine Learning Research Scientist: Generative Modeling for Planning

Nuro Nuro · Robotics · CA · Autonomy

Machine Learning Research Scientist focused on developing and scaling state-of-the-art generative models, particularly diffusion architectures, flow-matching, and energy-based models, for autonomous plan generation in robotics. The role involves leveraging foundation models (LLMs, world models), optimizing models with reinforcement learning and reward modeling, and developing controllable generative models. Collaboration across autonomy teams and deployment of models on NuroDriver are key aspects.

What you'd actually do

  1. Develop and scale state-of-the-art generative models—especially diffusion architectures, flow-matching techniques, and energy-based models —for autonomous plan generation.
  2. Build generative models with foundation models. Leverage large language models and world foundation models for reasoning, decision making and multi-modality generation.
  3. Optimize generative models using reinforcement learning to improve interactive reasoning. Explore reward modeling/learned verifier using generative models. Explore joint prediction and planning and self-play. Leverage generative models for active learning and world modeling.
  4. Develop controllable generative models to guide the generation process towards desired goals, conditions and rewards.
  5. Collaborate across autonomy teams while developing holistic solutions to top autonomy challenges. Understand issues, propose ideas, prioritize work and develop solutions to solve them, evaluate your solution by deploying the models on to the NuroDriver.

Skills

Required

  • Python
  • C++
  • generative models
  • diffusion models
  • flow matching
  • energy-based models
  • reinforcement learning
  • foundation models
  • LLMs
  • world foundation models

Nice to have

  • manipulation
  • path planning
  • autonomous driving
  • vision-language-action models
  • video generation
  • text-to-image generation
  • diffusion models for LLMs
  • self-play
  • active learning
  • world modeling

What the JD emphasized

  • Ph.D. (preferable) or M.Sc. with 2-3 years of experience
  • Demonstrated research publications in top conferences

Other signals

  • develop and scale state-of-the-art generative models
  • autonomous plan generation
  • foundation models
  • reinforcement learning
  • controllable generative models
  • deploying the models on to the NuroDriver