Member of Technical Staff - Simulation, Frontier AI Robotics

Amazon Amazon · Big Tech · San Francisco, CA · Software Development

The role focuses on developing 3D physics-based and photorealistic simulations for training large-scale machine learning models in robotics. This involves creating simulations for reinforcement learning, generating synthetic data, implementing robotics features, and building real-to-sim workflows to minimize sim-to-real gaps. The goal is to support the training of foundation models for robotics.

What you'd actually do

  1. Develop simulations for reinforcement learning, closed-loop simulations and synthetic data generation
  2. Implement essential robotics features, including accurate modeling of sensors, actuators, and controllers
  3. Build real-to-sim workflows for dynamic environments and robotics tasks
  4. Implement simulation features to minimize sim-to-real gaps through domain randomization and system identification
  5. Collaborate closely with a team of ML researchers to enable large-scale robotics training pipelines

Skills

Required

  • Python
  • C++
  • CUDA programming
  • TensorRT or similar ML optimization frameworks
  • Ability to optimize ML models for production
  • 5+ years of non-internship professional software development experience
  • 5+ years of leading design or architecture (design patterns, reliability and scaling) of new and existing systems experience
  • 5+ years of programming with at least one software programming language experience
  • Experience as a mentor, tech lead or leading an engineering team
  • Bachelor's degree in computer science or equivalent

Nice to have

  • Expertise in NVIDIA's ML stack (cuDNN, CUDA Graph, etc.)
  • Experience with ML compilers (ONNX Runtime, TVM, etc.)
  • Experience with transformer model optimization
  • Background in performance profiling and optimization
  • Experience working directly with research teams
  • Ability to build robust monitoring systems
  • Experience with large-scale ML serving systems

What the JD emphasized

  • foundation models for robotics
  • reinforcement learning
  • synthetic data generation
  • real-to-sim workflows
  • sim-to-real gaps

Other signals

  • foundation models for robotics
  • reinforcement learning
  • synthetic data generation
  • real-to-sim workflows
  • sim-to-real gaps