AI Infrastructure Engineer, Digital Optimus

Digital Optimus is the software counterpart to our physical humanoid, designed to interact with computer interfaces and perform long-horizon agentic behaviors. Our approach is modeled after real-time control policies rather than screenshot-based VLM agents, with the larger goal of integrating with Tesla's broader AI ecosystem. We're seeking an ML/RL Infra Engineer to build scalable, reliable infrastructure that powers these agents and enables seamless, high-volume rollouts for model evaluation & RL training. Top candidates will have deep experience in large-scale ML systems, high-performance training, and edge deployment, though evidence of exceptional ability matters more than relevance alone.

Responsibilities

Design & implement scalable distributed training infrastructure for large agentic models, supporting imitation learning, reinforcement learning (online & offline), and long-horizon training workflows Build high-fidelity, ultra-realistic training & simulation environments capable of handling complex, interruptible, long-context agent trajectories at massive scale Optimize ML and RL training pipelines for throughput, cost-efficiency, and reliability across multi-node GPU clusters Implement advanced model serving, quantization, distillation, and deployment strategies tailored for Tesla’s hardware platforms Collaborate with research, AI engineering, and production teams to productionize agent systems and integrate them with Tesla’s autonomy (FSD) and robotics (Optimus) platforms Design systems for efficient context management, checkpointing, and orchestration of long-horizon agentic workloads Continuously improve developer velocity through better tooling, CI/CD for ML, experiment tracking, and reproducible training environments

Requirements

Experience in ML infrastructure, large-scale distributed systems, or high-performance computing for deep learning/reinforcement learning Strong expertise with training frameworks (PyTorch, JAX, DeepSpeed, FSDP, Megatron, etc.) and distributed training at scale Deep knowledge of GPU/accelerator optimization, model parallelism, quantization, and edge deployment Proficiency in Python, Kubernetes, cloud infrastructure (or on-prem clusters), and modern MLOps practices Experience building data pipelines and simulation environments for reinforcement learning or robotics applications is highly valued Strong software engineering fundamentals, system design skills, and a passion for building reliable, observable, and high-performance ML platforms Ability to work effectively in a fast-paced, cross-functional environment with researchers and engineers