Computational Scientist / Scientific Machine Learning

Fusionality; Ecublens; On-site; CHF 120,000 - 190,000; Full time

Benefits offered by this job

Equity compensation

Competitive salary

Immigration assistance

Job summary

Fusionality in Lausanne is building a cutting-edge measurement, simulation, and control stack for future fusion reactors. We seek a PhD-level or senior Computational Scientist specializing in Scientific Machine Learning (SciML) to develop differentiable solvers that run against real plasma data and integrate into real control systems.

Role requires deep experience with differentiable frameworks (JAX, Julia, PyTorch) and strong numerical methods for ODEs/PDEs.

Qualifications

• Substantial code written in a differentiable framework. Our stack is in JAX, Julia or PyTorch experts are very welcome if they’re happy to switch.
• Strong background in numerical methods for ODEs and PDEs: discretisation, stiff and implicit solvers, nonlinear and sparse linear solvers.
• Demonstrated ability to operate at the frontier of SciML, Universal Differential Equations, Neural ODEs, and other emerging hybrid data-driven/mechanistic paradigms.
• Evidence of independent work at the highest level: a PhD or thesis, published code that is used by others, or an industry project you owned.
• Experience in the following fields an advantage but not strictly required: The JAX ecosystem: Diffrax, Equinox, Optax, Lineax.
• Plasma, fluid, electromagnetic, or other multiphysics simulation.
• Having led the design of a scientific codebase used by others.

Responsibilities

• Design, implement and validate differentiable solvers in JAX for plasma physics problems such as free-boundary equilibrium, transport and state estimation.
• Make them robust and fast enough for production use via advanced numerical techniques for differential equation time stepping, as well as optimization of GPU/TPU performance.
• Using automatic differentiation for model calibration against experimental data, parameter estimation and controller design and optimization, working with our data and control engineers.
• Code ownership: quality, testing, verification against reference solutions, packaging and documentation.

Skills

SciML

JAX

PyTorch

Differentiable solvers

GPU optimization

ODE methods

Education

PhD level

Tools

Diffrax

Equinox

Optax

Lineax

Job description

Fusionality is building the measurement, simulation, and control stack for the fusion energy reactors of the future.

Spun out of EPFL’s Swiss Plasma Center and backed by top investors, our technology enables reliable operation, real-time monitoring, and analysis of high-temperature plasmas in fusion devices. We are founded by world-class experts at the intersection of fusion physics, advanced control, and AI (ex-Google DeepMind, EPFL). We are assembling a founding engineering team in Lausanne to build our core technology platform.

We are looking for a Computational Scientist specialized in Scientific Machine Learning (SciML) to join our team. This job is for you if you know how to write numerical solvers that are fast, correct and differentiable end to end, and want to see them run against real plasma data and inside real control systems.

We are looking for a PhD-level or senior computational researcher for this role.

Responsibilities

Design, implement and validate differentiable solvers in JAX for plasma physics problems such as free-boundary equilibrium, transport and state estimation.

Make them robust and fast enough for production use via advanced numerical techniques for differential equation time stepping, as well as optimization of GPU/TPU performance.

Using automatic differentiation for model calibration against experimental data, parameter estimation and controller design and optimization, working with our data and control engineers.

Code ownership: quality, testing, verification against reference solutions, packaging and documentation.

Requirements

Substantial code written in a differentiable framework. Our stack is in JAX, Julia or PyTorch experts are very welcome if they’re happy to switch.

Strong background in numerical methods for ODEs and PDEs: discretisation, stiff and implicit solvers, nonlinear and sparse linear solvers.

Demonstrated ability to operate at the frontier of SciML, Universal Differential Equations, Neural ODEs, and other emerging hybrid data-driven/mechanistic paradigms.

Evidence of independent work at the highest level: a PhD or thesis, published code that is used by others, or an industry project you owned.

Experience in the following fields an advantage but not strictly required:

The JAX ecosystem: Diffrax, Equinox, Optax, Lineax.

Plasma, fluid, electromagnetic, or other multiphysics simulation.

Having led the design of a scientific codebase used by others.

What we offer

Play a key role at a dynamic, fast-growing startup, shaping the technical solutions and products from the ground up.

Build cutting-edge technology to monitor and control plasmas in real fusion devices, working side-by-side with leading plasma and control experts.

Great working conditions, located in Lausanne, Switzerland, a city and country known for its high quality of life.

Competitive salary and equity compensation and benefits including immigration assistance.