Computational Scientist, Differentiable Physics
Periodic Labs · Menlo Park, CA
Extracted from the posting
Builds differentiable, accelerator-ready solvers for continuum-physics problems, with emphasis on fluid dynamics and multiphysics. The work combines simulation with deep learning for surrogate modeling and inverse problems, using JAX or PyTorch with automatic differentiation. It also validates models against experiments and benchmarks and creates datasets for LLM development.
About the Role
Periodic Labs is building AI systems that can simulate physical science, verify predictions, and train on the full scientific method. We are looking for a Computational Scientist to build differentiable, accelerator-ready simulations for industrially relevant continuum-physics problems.
You should be equally comfortable with governing equations, solver code, and deep learning. We are open to expertise in any area of continuum-physics, with at least some experience in fluid dynamics. You will work on building simulation capabilities in challenging, data-limited domains requiring a mix of physics-based and empirical approaches.
What You’ll Do
- Build and extend differentiable solvers for continuum simulation (including but not limited to fluid dynamics), especially multi-scale and multi-physics problems.
- Implement numerical methods from equations and papers, and diagnose convergence, stability, and modeling failures.
- Combine simulation with deep learning for surrogate modeling, learned physics, inverse problems, parameter estimation, and optimization.
- Use automatic differentiation and modern accelerators with JAX or PyTorch to make simulations scalable and trainable.
- Validate models against experiments, trusted benchmarks, or high-fidelity simulations.
- Create datasets and evaluations to guide the development of LLMs to accelerate and automate these tasks.
You Will Thrive Here If You Have
- A PhD or equivalent research experience in applied mathematics, computational science, physics, engineering, computer science, or a related field.
- Code-level experience building or substantially modifying PDE solvers, numerical methods, or differentiable simulations.
- Deep expertise in at least one continuum domain, with breadth across domains or a demonstrated ability to learn new physics quickly.
- Meaningful experience building, training, and evaluating deep-learning models for physical systems.
- Strong Python and software-engineering skills, especially JAX, PyTorch, Julia, or C++.
- Experience applying simulation to realistic scientific or engineering problems, not only clean academic benchmarks.
- A startup mentality: ownership, good judgment under uncertainty, and enthusiasm for building from scratch.
Strong Candidates May Also Have
- Experience with fluid dynamics plus another continuum domain, or with multiphysics and multiscale modeling.
- Expertise in adjoint methods, implicit differentiation, differentiable programming, or scientific optimization.
- Experience accelerating scientific software on GPUs or TPUs.
- Contributions to scientific open-source software used by others.
- Experience connecting simulation to experiments, engineering decisions, semiconductors, or autonomous workflows.
Mechanics
- Minimum education: Bachelor's degree or similar experience
- Location: Menlo Park, CA
- Compensation: $250,000-350,000 + equity
- Visa sponsorship: Yes, we sponsor visas and will do everything we can to assist in this process.
