Machine Learning Engineer Intern - Planning

Team: Planning

Location: Santa Clara, CA

Commitment: Intern

Workplace Type: onsite

Salary:

Responsibilities:

• Multimodal Architecture Design: Modify and expand an existing Scene Transformer framework to generate a diverse set of ego-vehicle trajectories for ambiguous scenarios.
• Loss Function Engineering: Formulate and experiment with recursive, autoregressive, and multi-step losses to stabilize long-horizon trajectory rollouts.
• Diversity & Feasibility Optimization: Mitigate mode collapse to ensure the generated paths are distinctly varied yet kinematically feasible.
• Evaluation & Benchmarking: Design metrics to evaluate trajectory diversity, safety, and long-horizon accuracy against real-world driving logs.

Required Skills:

• Expertise in trajectory prediction, behavior forecasting, and multi-agent motion modeling
• Experience training large-scale models with self-supervision, contrastive learning, representation learning, or scene-level modeling
• Practical experience with PyTorch, PyTorch Lightning or JAX.
• Experience with uncertainty estimation (aleatoric, epistemic), multi-modal prediction, and probabilistic modeling
• Understanding of AV stack components: localization, perception, tracking, prediction, and planning.
• Strong debugging, experiment design, data analysis, and failure-mode investigation skills.