AI Engineer, World Modeling & Video Generation, Tesla AI

Tesla Tesla · Auto · Palo Alto, CA · Tesla AI

AI Engineer focused on training world models and video generation for robotics, emphasizing causal and physics-aware architectures, closed-loop reinforcement learning, and large-scale multimodal training with real-time inference.

What you'd actually do

  1. Design and train action-conditioned video generation models that predict future frames and sensor states
  2. Develop causal, physics-aware architectures that model interactions, motion, and environmental dynamics
  3. Integrate 3D generative techniques such as Gaussian Splatting and volumetric rendering for high-fidelity realism
  4. Implement closed-loop training systems where models iteratively refine predictions through feedback and simulation
  5. Optimize distributed pipelines for large-scale multimodal training and real-time inference

Skills

Required

  • Expertise in generative video or world model architectures
  • Strong background in spatiotemporal modeling, 3D scene understanding, or neural simulation
  • Proficiency with PyTorch or JAX
  • Experience in large-scale distributed training, especially the different forms of parallelism
  • Familiarity with reinforcement or imitation learning in simulated or embodied environments

Nice to have

  • Curiosity about building intelligent systems that understand and generate the world around them

What the JD emphasized

  • action-conditioned video generation models
  • causal, physics-aware architectures
  • closed-loop training systems
  • large-scale multimodal training
  • real-time inference
  • generative video or world model architectures
  • spatiotemporal modeling
  • 3D scene understanding
  • neural simulation
  • large-scale distributed training
  • reinforcement or imitation learning

Other signals

  • training world models
  • action-conditioned video generation
  • causal, physics-aware architectures
  • closed-loop reinforcement learning
  • large-scale multimodal training
  • real-time inference