PhD ML Intern: Controllable Generative World Models

Waymo LLC; Mountain View (WY); Híbrido; USD 97.000 - 138.000; Jornada completa

Ventajas ofrecidas por este puesto de trabajo

Benefits programs

Hybrid onsite internship

Descripción de la vacante

Waymo LLC is seeking a PhD candidate for a hybrid onsite internship focused on autonomous driving software research. You will work on controllable generation, world-modeling, and multimodal simulation with teams across Sim Realism and Foundation Models.

The role requires strong Python/JAX/PyTorch and experience with Transformers, diffusion, and multi-agent simulation. Applicants should expect to split time between research and implementation.

Formación

• Pursuing a PhD with anticipated graduation between Dec 2027 and Aug 2028 and returning to a degree program after internship.
• Strong tensor implementation and debugging in Python and JAX/Flax or PyTorch with hands-on neural architectures.
• Experience with Transformers, Diffusion/Flow-Matching models, or deep RL/critic networks.

Responsabilidades

• Design, implement, and scale controllable generation and test-time compute algorithms for world models.
• Investigate multimodal sensor conditioning and representation alignment for world simulation.
• Leverage realism critics and large-scale TPU workflows to evaluate long-tail scenarios.
• Write clean, modular JAX code in core simulation and world-modeling codebases with researchers.

Conocimientos

Python

JAX/Flax

PyTorch

Transformers

Diffusion models

Deep RL

Educación

PhD in Computer Science, ML, Robotics, or related field

Herramientas

TPU clusters

Descripción del empleo

Waymo LLC is seeking a PhD candidate for a hybrid onsite internship focused on autonomous driving software research. You will work on controllable generation, world-modeling, and multimodal simulation with teams across Sim Realism and Foundation Models.

The role requires strong Python/JAX/PyTorch and experience with Transformers, diffusion, and multi-agent simulation. Applicants should expect to split time between research and implementation.