A Scientific Machine Learning Researcher develops machine learning methods that respect the laws of physics. The field, often shortened to SciML, sits between deep learning, numerical analysis and the physical sciences. Its researchers build models that can learn from simulation and experimental data while staying consistent with conservation laws, symmetries and known equations. The role is also advertised as Research Scientist (Physics AI), AI Research Scientist (Simulation), Machine Learning Researcher (Physical Sciences), Computational Physicist (Machine Learning), Applied Scientist (Physics-Based AI) and Research Engineer (AI for Science).
Organisations hire these researchers when off-the-shelf machine learning isn’t good enough for physical problems. A standard neural network can fit data well and still predict something physically impossible, or fail badly outside its training range. SciML researchers design architectures, loss functions and training methods that generalise better, need less data and give engineers and scientists results they can trust.
What does a Scientific Machine Learning Researcher do?
The role is research-led, with a clear line to application. Researchers define open problems, develop and test new methods, and work with simulation engineers or domain scientists to apply them to real systems.
- Developing neural operators, including Fourier neural operators and DeepONet-style architectures, that learn mappings between functions rather than fixed inputs
- Building physics-informed neural networks and other methods that build governing equations into training
- Designing graph neural networks and transformer models for mesh, particle and field data
- Researching differentiable simulation, hybrid solver and ML approaches, and learned closures for turbulence and other sub-grid physics
- Quantifying uncertainty and testing how models generalise to new geometries, conditions and regimes
- Creating benchmark datasets and evaluation protocols that reflect real scientific and engineering use
- Publishing research and turning prototypes into tools that applied teams can adopt
Key skills and backgrounds
Most candidates hold a PhD in applied mathematics, physics, computational science, mechanical or aerospace engineering, or machine learning with a strong physical sciences focus. Many come from computational physics or numerical methods and moved into ML during or after their doctorate. Senior researchers usually have a publication record at machine learning or computational science venues and a clear research agenda.
Look for people who can explain why a method works, not just that it does. Ask candidates where their favourite approach breaks down and how they would detect that in practice.
- Strong mathematical foundations in partial differential equations, numerical analysis and optimisation
- Deep learning research experience in PyTorch or JAX, including custom architectures and training procedures
- Knowledge of classical numerical methods, like finite volume, finite element and spectral methods
- Neural operators, physics-informed learning, graph neural networks and equivariant models
- Uncertainty quantification, Bayesian methods and surrogate modelling
- High performance computing and GPU programming for large training runs and datasets
- Clear scientific writing and the ability to work with domain experts outside ML
Where Scientific Machine Learning Researchers work
Industrial research labs in automotive, aerospace, energy and semiconductors hire SciML researchers to speed up design and simulation. Technology companies and AI labs run AI for science groups working on weather, climate, materials and molecular modelling. Simulation software vendors build research teams to bring ML into their products. National laboratories, weather and climate agencies, research institutes and universities also hire for the role, and many researchers move between academia and industry.
Scientific Machine Learning Researcher vs related roles
Several research roles overlap, so the distinctions help when writing a job description.
- A Simulation Engineer (AI/ML) applies proven ML methods to engineering simulation problems, where a SciML researcher develops the new methods those engineers later use.
- A Research Scientist in general machine learning works on methods across domains, usually without the numerical analysis and physics background this role depends on.
- A Computational Physicist traditionally builds and runs numerical simulations, and moves into this role when machine learning becomes a central part of the work.
Salary and market data
Pay for SciML researchers depends on seniority, publication record and whether the employer is an industrial lab, a technology company or a public research body. Our AI and deep tech salary guides cover research and ML roles, and the deep tech salary calculator helps you benchmark a package.
Hiring Scientific Machine Learning Researchers with DeepRec
DeepRec recruits for AI4Science and AI research teams across Europe and the US, from offices in London, Berlin, Zurich, Dublin and Boston. Our consultants screen technically, discussing candidates’ methods, the problems they’ve applied them to and how they validated results. Many strong researchers are in academia or national labs and not actively looking, so we search through research networks as well as applications. We work on a permanent, contract, retained and executive search basis. To start a search, visit Looking to Hire.