Foundation models for scientific and industrial data.
Expert AI: foundation models for the data that runs science and industry
We define Expert AI as foundation models built for scientific and industrial data: continuous and irregular time signals, spatiotemporal fields, multimodal scientific recordings, time-stamped event logs from real organizations. The kind of data that doesn’t look like text, doesn’t fit transformer assumptions, and rarely leaves the institutions that hold it. This isn’t a fine-tuning problem. It requires architectures designed for the data, infrastructure designed for how the data lives, and a tech stack of AI components purpose-built to assemble foundation models for new domains.
General-purpose AI was never built for science and industry. The data is sparse, noisy, distributed, and governed by strict rules. It demands expert intelligence: models that understand the structure of a domain, not just the surface of its data.
Our approach is co-foundation with domain experts: AI and domain science evolving together. We build domain-specialised foundation models trained to master a field’s data, constraints, and logic from the ground up. We work hand in hand with the scientists, clinicians, and engineers who know the domain best. Our models learn from distributed sources through decentralised architectures, keeping sensitive data where it belongs.
We anchor in brain and neuroimaging, the most complex signal of all and the hardest class of expert data we know how to characterize. In parallel, our cross-domain transfer programme works on extending the same components and methods to open new application domains.
We invest in demonstrators that turn research into tangible proof, working with partners on concrete use cases that test our models in real conditions and inform the next iteration of the science.
Our team finished 2nd out of 1,183 in the NeurIPS 2025 EEG Challenge, and we publish at Tier-1 AI conferencesincluding NeurIPS, ICML, and CVPR. This dual focus, rigorous science and applied impact, defines our approach.
This role sits at the intersection of numerical physics / simulation, physics-informed ML, and surrogate modeling with a strong pull toward high‑impact industrial domains such as aeronautics, automotive, and non‑destructive testing (NDT).
We are open in how we frame the position (Research Scientist vs Applied Scientist): what matters is a profile that can do serious research while keeping a clear path to deployment and operational impact, including collaborations with labs and industrial partners when relevant.
What you’ll work on
Physics-informed & surrogate modeling (PINNs / Neural Operators / Neural PDEs)
Build fast, accurate surrogates to emulate expensive simulations or physical processes (e.g., PDE-driven systems, complex boundary conditions, multi-physics settings).
Explore approaches such as Fourier Neural Operators, operator learning, PINNs, and hybrid ML+numerics methods.
Computational / numerical physics meets foundation models
Help us extend “foundation model thinking” to physical domains: pretraining, adaptation, and evaluation for scientific/industrial data (fields, meshes, sensor arrays, tomography/ultrasound-like signals, etc.).
Develop strategies that work under scarce, noisy, irregular, or biased physical datasets.
Architecture, inductive biases, and messy reality
Design architectures that respect physical structure: invariances/equivariance, geometry, irregular spatial/temporal grids, mesh-based data, multimodal measurement pipelines.
Investigate failure modes: when deep learning breaks in physics settings, why it breaks, and what to do about it.
From research to demonstrators
Own projects end-to-end: hypotheses → experiments → training on GPU infrastructure → robust evaluation → prototype integration.
Define evaluation protocols aligned with constraints like calibration/uncertainty, robustness, traceability, and cost of error.
Education / experience: PhD (preferred) or Master’s with strong research/applied experience in Computational Physics, Numerical Physics, Scientific Computing, Machine Learning, or a closely related quantitative field.
Physics + ML depth: strong grounding in physical modeling and numerical methods (e.g., PDEs, boundary conditions, discretization, simulation workflows) plus solid deep learning fundamentals.
Research discipline: ability to design rigorous experiments, debug systematically, and produce clear, defensible conclusions.
Engineering maturity: strong Python skills; experience with PyTorch and training on GPU(s)/clusters
Nice to have
Experience with Neural Operators, surrogate modeling, PINNs, geometric deep learning, mesh/point cloud learning, or scientific ML toolchains.
Domain exposure to aeronautics, automotive, or industrial R&D environments
Experience with inverse problems, uncertainty quantification, calibration, or safety/robustness constraints in scientific/industrial ML.
Publications or open artifacts in ML/physics venues (NeurIPS / ICML / ICLR; NeurIPS ML & Physical Sciences workshop; relevant physics/engineering journals/conferences).
Recruitment prescreen (remote 30-45min)
Scientific deep dive (remote-45min)
Half-day of scientific interviews (Architecture - Coding - Research talk) + Culture fit
References call