About the Role
You will develop efficient generative world models and planners for closed-loop driving simulation. You will optimize inference performance, define fidelity and coherence metrics, run experiments, integrate models into training and evaluation, measure sim-to-real performance, and mentor junior researchers.
Requirements
•4+ years of machine learning research or engineering experience focused on generative video or world models
•Knowledge of diffusion and latent-video models
•Experience improving sampling efficiency or model throughput
•Experience with high-dimensional temporal or spatiotemporal data
•Python and PyTorch engineering skills
•Experience building research-grade production tools
•Publication record or contributions to open-source machine-learning tooling
Responsibilities
•Develop efficient generative world models with real-time rollouts and controllable scene editing
•Architect interactive world models for reinforcement learning, planning, and safety evaluation
•Optimize latent compression, context pruning, and inference latency
•Define metrics for long-horizon coherence, physics fidelity, and planner integration
•Run ablation and scaling studies
•Integrate models into closed-loop training and evaluation
•Measure sim-to-real gaps against on-road driving-model results
•Mentor junior researchers and shape technical roadmaps
Benefits
•Hybrid working
•Core working hours with schedule flexibility
