Principal Machine Learning Engineer, Embodied AI and Smart Npcs

Roblox Roblox · Consumer · San Mateo, CA · Machine Learning

Principal ML Engineer focused on Embodied AI and Behavioral Agents for Roblox's 3D gaming platform. The role involves research and development of agents with human-like movement and strategic reasoning, leveraging Imitation Learning, Reinforcement Learning, Computer Vision, Computer Graphics, Robotics, and Agentic Reasoning. The goal is to create generalizable agents that can perceive 3D environments, understand game rules, plan long-term strategies, and execute complex physics-based actions, impacting millions of users in real-time gaming environments.

What you'd actually do

  1. Design and implement foundation models for embodied agents that master both fluid, human-like movement and high-level strategic reasoning.
  2. Define the long-term roadmap for Game AI and Embodied Intelligence, acting as a technical bar-raiser for code quality and architectural design.
  3. Balance the exploration of cutting-edge deep learning research with the practical constraints of serving models to millions of concurrent users.
  4. Mentor fellow engineers and researchers, fostering a culture of technical excellence and scientific inquiry.
  5. Collaborate with Product Managers, Backend and Game Engine Engineers and other Roblox team members.

Skills

Required

  • Python
  • PyTorch/TensorFlow
  • Imitation Learning
  • Robotics
  • Reinforcement Learning
  • Computer Graphics
  • Computer Vision
  • low-latency motor control
  • high-level strategic reasoning
  • training models on large-scale distributed clusters
  • inference in real-time gaming environments

Nice to have

  • PhD or Master’s in Computer Science, Applied Math, or other field
  • record of top-tier publications (e.g., NeurIPS, ICML, CVPR, AAAI, SIGGRAPH, etc) in embodied agents or related domains
  • C#
  • C++
  • systems languages

What the JD emphasized

  • applying research to tangible products
  • experience training models on large-scale distributed clusters and understanding the challenges of inference in real-time gaming environments
  • moving beyond academic benchmarks to launch scalable solutions that directly impact millions of users

Other signals

  • building agents capable of complex 3D gameplay
  • unblocking many use cases across Roblox
  • research and development of Embodied AI and Behavioral Agents
  • revolutionize how games are created and played on Roblox
  • bridge the gap between cutting-edge research and massive-scale product application