ML Engineer, Surrogate Modeling (Vehicle Engineering)

SpaceX•Hawthorne, CA

16d

About The Position

Be a member of the AI for Vehicle Engineering team, focusing on developing high-performance surrogate models to solve complex physics and engineering problems for our launch vehicles and spacecraft. Our team builds AI systems that accelerate engineering analysis, simulation, development, testing, avionics design, flight data review, logistics, and mission operations. Your work will directly support the world’s largest communication and AI satellite constellations, accelerate rapid reuse of the Falcon launch vehicle, and contribute to the development of the world’s largest rocket capable of sending humans to Mars. In this role, you will develop, train, tune, and deploy AI surrogate models to dramatically accelerate engineering simulations, including but not limited to FEA, CFD, thermal, and structural analysis. You will work closely with hardware, simulation, and domain engineers to build these systems from the ground up. You will leverage state-of-the-art surrogate modeling techniques and create novel methodologies that push the frontier of what is possible in ML for physics while tackling real-world problems. Aerospace experience is not required. We are looking for smart, motivated, collaborative engineers who love applying machine learning to hard scientific problems and want to make a direct impact on SpaceX’s mission.

Requirements

Bachelor’s degree in computer science, data science, engineering, math, physics, or a related technical discipline; OR 4+ years of professional experience building software in lieu of a degree
1+ years of software development experience in Python for machine learning, AI, or data science applications

Nice To Haves

Master’s or PhD in computer science, machine learning, engineering, or a related field with a focus on surrogate modeling or AI for scientific/engineering simulation
Demonstrated experience training, tuning, and deploying production-grade ML surrogate models in real engineering workflows
Expert-level understanding of at least one modern architecture class such as Fourier Neural Operators (FNO), neural operators, MeshGraphNet, Transolver, graph neural networks, physics-informed neural networks, or other surrogate model architecture
Experience solving inverse problems such as geometry optimization or design under uncertainty
Strong understanding of traditional simulation and numerical methods (CFD, FEA, thermal analysis, etc) and how to integrate them with surrogate models
Experience with uncertainty quantification techniques for surrogate models
Hands-on experience building active learning or adaptive sampling pipelines
Proficiency with deep learning frameworks such as PyTorch, TensorFlow, or JAX
Experience with surrogate modeling libraries such as NVIDIA PhysicsNemo or similar
Experience developing on Linux systems with GPU accelerators
Strong understanding of software engineering best practices including version control, testing, and continuous integration
Solid foundation in statistics, numerical methods, and core machine learning algorithms

Responsibilities

Develop, train, evaluate, and deploy production-grade AI surrogate models that accelerate critical engineering simulation workflows
Design and implement State-of-the-Art (SOTA) neural architectures and training strategies tailored to complex engineering problem domains
Build scalable data pipelines to preprocess, manage, and utilize tens of thousands of high-fidelity simulation results
Stay current with the latest research in neural operators, physics-informed ML, and surrogate modeling, implementing new techniques when needed
Collaborate with peers on architecture, design, and code reviews
Deep dive into engineering problems to identify where AI can deliver the highest leverage and most reliable solutions
Develop and apply techniques for uncertainty quantification, active learning, and inverse problems (e.g., geometry and shape optimization)
Ensure all AI systems are rigorously validated and vetted for accuracy, robustness, and reliability in engineering use