AI Research Intern – Predictive World Model

The role

You will drive a focused research project on predictive world models that learn how the physical world changes under an agent's actions. The work supports simulation and training of driving and robotic policies, using generative architectures and large-scale multimodal data.

What you will do

• Drive a focused research project on predictive world models, from problem formulation through evaluation, with a mentor and the research team.
• Contribute to multi-view future prediction and generation, including action-conditioned and action-free forecasting.
• Work on architectures where one backbone both predicts the future and produces trajectories or actions.
• Apply predictive pre-training to improve Vision-Language-Action (VLA) driving performance.
• Extend prediction into a shared multimodal latent space that includes 3D scene representations, occupancy and reward signals.
• Investigate cross-embodiment generalization so one world model transfers across vehicles, robots and sensor setups.
• Build evaluation methods for predictive world models, including physical plausibility and closed-loop policy performance.
• Work with research engineers to move prototypes into scalable training and inference pipelines, and publish or open-source results where appropriate.

Required qualifications

• Currently pursuing a PhD in Engineering, Computer Science or a related field, focused on Deep Learning, Computer Vision or Generative Models.
• First-author publications at top-tier venues such as CVPR, ICCV, ECCV, NeurIPS, ICLR, ICML, CoRL, RSS or SIGGRAPH. Work under submission may be presented as an arXiv preprint.
• Strong, up-to-date foundation in generative modeling and experimental methodology.
• Hands-on experience building, training, fine-tuning and evaluating models in PyTorch or JAX.
• Strong Python programming and software design skills, including data structures, algorithms, code optimization and large-scale data processing.
• Available to commit to a minimum of 12 weeks and work on-site at the Santa Clara office.

Nice to have

• Hands-on experience with generative models for video or 3D, such as diffusion, flow matching, autoregressive video prediction, NeRF or Gaussian Splatting.
• Experience with world models or learned simulators for decision making, including model-based reinforcement learning and VLA models.
• Experience with multimodal foundation models and video tokenizers or VAEs, including pretraining or adapting large pretrained backbones.
• Prior research internship experience in autonomous driving, robotics or embodied AI, or contributions to widely used open-source projects.

Benefits

• A fun, supportive and engaging environment.
• Infrastructure and computational resources to support your work.
• Opportunity to work on cutting-edge technology with top talent in the field.
• Opportunity to make significant impact on the transportation revolution by advancing autonomous driving.
• Competitive compensation package.
• Snacks, lunches, dinners and fun activities.

Before you apply

• Strong results are supported for publication at top-tier venues.

About the company

XPENG is a smart technology company that integrates AI and autonomous driving into electric vehicles, eVTOL aircraft and robotics. It focuses on intelligent mobility and R&D in AI, machine learning and smart connectivity.