Differentiable Physics Scientist - PDEs, ML & Acceleration

Periodic Labs; Menlo Park (CA); On-site; USD 250,000 - 350,000; Full time

Job summary

Periodic Labs is seeking a Computational Scientist to build differentiable, accelerator-ready simulations for continuum-physics problems. You will work across governing equations, solver code, and deep learning to tackle multi-scale, multi-physics challenges in data-limited domains.

You will extend solvers, implement numerical methods, and combine physics-based models with deep learning for surrogate modeling and optimization. Strong Python/JAX/PyTorch and GPU/TPU experience preferred.

Qualifications

• PhD or equivalent research experience in a related field.
• Code-level experience building or modifying PDE solvers or differentiable simulations.
• Deep expertise in at least one continuum domain.
• Experience building and evaluating deep-learning models for physical systems.

Responsibilities

• Build and extend differentiable solvers for continuum simulation, including 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 accelerators with JAX or PyTorch to scale simulations and enable training.
• Validate models against experiments, trusted benchmarks, or high-fidelity simulations.
• Create datasets and evaluations to guide development of LLMs to accelerate these tasks.

Skills

Python

JAX

PyTorch

Julia

C++

Software engineering

Differentiable simulations

Education

PhD or equivalent in applied mathematics / computational science / physics / engineering / CS

Bachelor's degree or equivalent

Tools

PDE solvers

GPU/TPU acceleration

Job description

Periodic Labs is seeking a Computational Scientist to build differentiable, accelerator-ready simulations for continuum-physics problems. You will work across governing equations, solver code, and deep learning to tackle multi-scale, multi-physics challenges in data-limited domains.

You will extend solvers, implement numerical methods, and combine physics-based models with deep learning for surrogate modeling and optimization. Strong Python/JAX/PyTorch and GPU/TPU experience preferred.