AI Engineer, Model Distillation, Tesla AI

About the role

Join the Tesla AI team to build intelligence for millions of robots by training frontier-scale foundation models. You will focus on transferring capabilities from massive teacher models to compact, efficient student models that operate under strict real-world constraints.

Key facts

What you'll do

• Create and run distillation pipelines to move reasoning, vision-language, and action policies from large teachers to student models for FSD, Optimus, and Digital Optimus.
• Develop distillation methods including logit-level, sequence-level, and trajectory-level techniques using synthetic data, soft-labels, SFT, and RL.
• Analyze scaling behaviors regarding teacher size, student capacity, compute allocation, and data mixtures to optimize on-robot performance.
• Engineer infrastructure for large-scale teacher inference, data curation, and distributed training while addressing memory and compute bottlenecks.
• Validate distilled models against teacher fidelity and production metrics to ensure performance and safety.
• Collaborate with cross-functional teams to deploy models into production environments.
• Build reusable tools and frameworks to support distillation across all Tesla AI model families.

Requirements

• Deep expertise in deep learning, specifically in compressing, training, or distilling large-scale multimodal, vision, or language models.
• Understanding of teacher-student dynamics and the balance between compute, latency, and capability.
• Knowledge of loss functions like KL or soft-targets and modern architectures such as hybrid attention or mixture of experts.
• Experience with post-training processes including RL, policy optimization, and supervised fine-tuning.
• Proficiency in distributed computing and large-scale training or inference systems.
• Strong Python programming skills and software engineering experience.
• Experience with JAX, TensorFlow, or PyTorch.
• Ability to troubleshoot system-level issues across training, data, and deployment.

Skills & tools

• Python
• PyTorch, TensorFlow, or JAX
• Distillation (Logit, sequence, trajectory)
• Distributed computing
• Supervised Fine-Tuning (SFT)
• Reinforcement Learning (RL)

Practical notes

Benefits begin on the first day of employment and include $0 payroll deduction medical, dental, and vision plans, HSA contributions, 401(k) with match, and stock purchase plans. Additional perks include family-building support, disability insurance, paid time off, and commuter benefits. Compensation is based on individual experience, skills, and market location.