Physics-Informed Neural Operators (PINO) for Ultra-Fast Tomography: Toward Fundamental and Generalizable AI for Industrial Fluid Mechanics
Horizon Europe funding secured

Location: fluiidd (La Ciotat, France) in collaboration with CEA and CNRS

Academic supervision: Guillaume Ricciardi (PhD, HDR), CEA and Cédric Bellis (PhD, HDR), CNRS

Industrial supervision: Mathieu Darnajou (PhD), CEO of fluiidd

Framework: Deep Tech Startup / Horizon Europe Marie Skłodowska-Curie Actions (MSCA)

1. Context and Challenges

A CEA spin-off, fluiidd embeds neural networks into its multiphysics tomography sensor to deliver predictive health monitoring of industrial flows. This new approach enables “seeing” and forecasting failures deep inside complex flows. These failures are currently identified relative to an arbitrary nominal state that does not account for the underlying physics. To improve model prediction and within the European COMBINE project, we are recruiting a PhD candidate to bridge fundamental physics and time-series characterization.

You will join an agile team composed of:

• A PhD student in AI/Control: focused on anomaly detection in time series.
• An MLOps Engineer: responsible for deployment and production of models.
• An Embedded Software Engineer: ensuring high-frequency data acquisition.

Your role: Be the team’s “physical brain.” You will turn raw electrical and vibrational signals into invariant physical quantities.

2. Scientific Missions and Objectives

The goal is to develop a physical “Rosetta Stone” for our sensors. Your work will focus on the motion of objects (e.g., assemblies, valves, blades) in flow:

• State Observer Development: Extract position (x,y) and vibration frequencies of a solid from conductivity measurements (EIT) and accelerometry.

• Transduction and Invariants: Convert these measurements into dimensionless quantities (Reynolds number, void fraction, cavitation intensity, etc.) to enable AI to generalize diagnostics to any industrial system. Exploration of PINN and PINO methods.

• Phenomenological Validation: Conduct experimental campaigns to validate fluid–structure coupling models, especially during critical phenomena (flow-induced vibrations, cavitation, clogging).

3. Desired Profile & Skills

We are looking for a high-level Computational Physicist with a passion for instrumentation.

Required technical skills:

• Physics: Strong background in fluid mechanics, electromagnetism, and inverse problems.
• Mathematics: Neural networks, numerical analysis, signal processing, basics of finite element methods (FEM).
• Programming: Proficiency in Python (scientific and AI libraries) and C++.
• Languages: Fluent English (European research context). French is a plus.

4. Eligibility (Strict Horizon Europe Criteria)

In accordance with EU mobility rules:

• Mobility Rule: Candidates must not have resided or carried out their main activity (work, studies) in France for more than 12 months in the 36 months prior to recruitment.

• Degree: Hold a Master’s degree (or equivalent) allowing PhD enrollment and must not already hold a PhD.

5. Conditions and Benefits

This position offers outstanding research conditions within the French Deep Tech ecosystem:

• Attractive salary: Defined by Horizon Europe scales (Living Allowance + Mobility Allowance + Family Allowance if applicable)

• Environment: Position based in La Ciotat (seaside), within an award-winning startup (Vivatech, Le Point Innovators).

• Network: Access to the European COMBINE network, international training, and collaborations with CEA and world-leading experts.

Why apply?

You won’t just write a PhD thesis; you will build the technological core of a company decarbonizing industry. You will be the one enabling AI to understand the physical world.

More info:

• https://cordis.europa.eu/project/id/101227547

• https://fluiidd.com/