Research Scientist - Generalization

About NPco

Engineering progress is constrained by the speed of simulation. In aerospace, energy, and other safety-critical industries, design cycles are still governed by solvers built for a different era, accurate, but slow and costly, and fundamentally incompatible with real-time decision-making. NPco exists to change that constraint.

Founded by Emmanuel and Matthieu, both former researchers at France's national institute for research in digital science and technology (Inria) working on AI powered simulation, NPco is building foundational AI models for physical simulation. By combining advances in large-scale model architectures with modern compute, we enable real-time iteration across fluid dynamics, structural mechanics, and electromagnetics, without sacrificing physical fidelity.

We want to push engineering towards real time feedback loops. We want Aircraft components to be optimized aerodynamically in real time. Wind farm layouts to be explored interactively to maximize energy yield. Thermal stress in nuclear infrastructure to be anticipated before it becomes a limiting factor. These will come as the natural consequence of the faster physics simulation we are building.

We are deeply embedded in the European research ecosystem, backed by engineers and founders from organizations including Mistral AI, Dataiku, and DeepMind. As we scale our technology into real-world industrial deployments, we are looking for engineers who want to work at the boundary between frontier AI and the most demanding physical systems on earth. We want to become the new technology layer of modern engineering. Our ambition is global, our first office is in Paris, France.

Your Mission

Strengthen generalization of our physics models across geometries, boundary conditions, and regimes, ensuring robust performance in new domains.

Generalization Research: Improve model generalization across geometries, boundary conditions, and regimes.

Inductive Biases: Research inductive biases (equivariance, invariances, regularization) for physics tasks.

OOD Testing: Build OOD benchmarks and stress tests to measure reliability in new domains.

Transfer Strategy: Define data-augmentation and curriculum strategies for robust transfer.

Education: You have a PhD or equivalent research experience and have spent significant time training large foundation models.

Programming: Advanced proficiency in Python and deep learning frameworks, specifically PyTorch.

Scientific Interest: A strong interest in physics or numerical simulation is a plus, even if your primary background is in pure AI.

Mindset: You are a builder who enjoys moving from theoretical research to functional, industrial-scale products.

Culture: Small team of committed top scientists and engineers in their field.

High Impact: Join a founding team where your work directly shapes the product and its deployment with global leaders.

Work on physical AI that delivers immediate, real-world value, making planes, windfarms, and nuclear plants more efficient and sustainable.

Frontier AI: Work at the intersection of Generative AI, High-Performance Computing, and complex physics.

Direct Influence: As one of our first Engineers, you will have a massive say in our roadmap and how our user interface evolves to meet engineering needs.

Momentum: Join us at a key turning point as we scale from technical validation to major commercial expansion.

‍

The job is published on NP Company’s website. Open the original posting ↗ to submit your application. Screeningpass does not fill in a form for you.

Start a contact search from this page. Screeningpass suggests up to three people connected to the role. Choose whom to write to, then Screeningpass looks for their professional email and prepares a message. You review it and send it yourself with Gmail. An email address is not always available, and a reply is never guaranteed.

A search costs 1 credit. Your credit is returned if the search fails. Signing up gives you 3 credits, enough for 3 searches.

Screeningpass brings together internships, apprenticeships and jobs published by companies. For a job, Screeningpass finds someone to contact at the company, looks for their professional email and prepares a message using your CV. See how it works in detail.