I am currently a research engineer in the CFL team at CEMEF, Sophia-Antipolis. My current work is concerned with the application of machine learning algorithms to CFD problems (see the MINDS project). Find out more about what I do below or by downloading my résumé. You can also find a list of the publications I was associated to here.
Research engineer at CEMEF (Mines ParisTech)
Topic: Machine learning for CFD applications
Research engineer at INRIA
Topic: Development of a discontinuous Galerkin-based software suite (see here)
PhD in applied mathematics and numerical simulation at INRIA
Topic: Discontinuous Galerkin time-domain methods for nano-optics (see here)
Engineering degree in applied mathematics at ENSTA ParisTech
Specialized in applied mathematics
I am working with P. Garnier and E. Hachem in the context of ERC Cure on graph neural network models for flow computation in intracranial aneurysms.
Beacon is a small library that proposes single-file flow control environments in python, to benchmark and tune DRL algorithms.
In the context of a benchmark library, I re-implemented an environment from Beintema et al. which aims at controlling the Rayleigh-Benard convection cell with bottom plate temperatures actuators.
For this paper, I re-implemented a CFD environment from Belus et al. based on the Shkadov equations, which aims at controlling the instabilities in a falling liquid film.
I wrote a small python-based DEM solver, available here.
Between 2015 and 2018, I have led the development of the Diogenes software suite. It is based on discontinuous Galerkin methods, and includes advanced features for the solving of nanophotonics problems in time- and frequency-domain.
During my PhD, I made a short numerical study on the effects of lithography defects on the behavior of reflectarrays (see here). Here, a non-cartesian doubly-periodic anisotropic metasurface is used to deflect a normally-incident plane wave with a non-cartesian angle.
I designed a control problem in which an agents learns to mix a passive scalar quantity in a square cavity using boundary velocities control.
I am working on a modular DRL library, in order to couple it with in-house solvers for flow control. More to come soon !
I wrote a small python-based LBM solver, available here.
I worked on the extension of the field transform technique to handle polychromatic plane wave analysis at oblique incidence with periodic boundary conditions in time-domain. Here, a doubly periodic transmittive color filter illuminated from below with a 30 degrees incidence angle. See related publication here.
I developed a specific source module able to take into account mode dispersion (index and field profile) to propagate polychromatic pulses in waveguides. The mode solution and the dispersion relation can be analytic, or provided through external files. Here, a TE01-to-TE02 mode converter is simulated on a 2 GHz large spectrum. See related publication here.
I have been working on various mesh algorithms in a new FE framework. Here, successive interface fittings of a pentagon on a base Delaunay mesh.
I reproduced a DRL-based flow control case from Rabault et al. for an upcoming technical paper. Here, the agent controls two small jets at the top and bottom of the cylinder.
I've recently been testing the PBO algorithm on packing problems, and obtained results similar to litterature for up to 20 degrees of freedom.
I developped an optimization technique based on policy gradient methods and evolution strategies. Here, it is used to design a control law for the Lorenz attractor. See related publication here.
The basis library of the Diogenes software suite is able to handle multi-element meshes composed of both hexahedra and tetrahedra (reading, neighbor-finding, partitionning, ...). Here, a hybrid mesh processed for the Diogenes-HDGFD solver.
A part of my PhD was dedicated to the study of isoparametric elements in nanophotonics computations (see here). Here, we show a resonance in the gap between two metallic nanospheres discretized with quadratic tetrahedra.