Bio
I am an R&D Engineer and PhD Candidate (expected Oct 2026) in computational cosmology at the AstroParticule & Cosmology Laboratory (APC, CNRS/IN2P3) and Université Paris Cité. I am advised by Eric Aubourg, Alexandre Boucaud, Josquin Errard, and François Lanusse. My thesis, “Differentiable Field-Level Cosmology: CMB, Weak Lensing, and GPU-Accelerated Inference,” develops the differentiable simulation and inference stack needed to extract cosmological information from the next generation of large-scale-structure surveys.
Interests
My research sits at the intersection of high-dimensional statistical modeling, differentiable programming, and large-scale distributed computing. The problems I find most interesting are those where the bottleneck is the engineering rather than the underlying physics — extracting principled posteriors from cosmological surveys requires forward models that are simultaneously physically accurate, automatically differentiable, and capable of scaling across thousands of accelerators. Most of my time is spent building the JAX-based simulation and inference tools that make this kind of analysis tractable.
Research
I develop differentiable N-body solvers — most notably JaxPM, which I maintain — capable of running at Rubin LSST scale. Multi-node domain decomposition and automatic differentiation make it possible to perform hierarchical Bayesian inference directly over forward simulations of large-scale structure, paving the way for full-field inference and near-optimal extraction of cosmological information.
As a member of the Simons Observatory, I also work on CMB component separation, focusing on the spatial variability of foreground spectral indices, which biases the tensor-to-scalar ratio if not modeled carefully. To this end, I develop the FURAX framework — a fast, scalable minimizer designed to remain robust under the extremely noisy regimes expected for upcoming full-sky missions like LiteBIRD.
Prior to academia
Before starting the PhD I spent five years in industry. I was a Software Infrastructure Engineer at Dassault Systèmes, where I worked inside the multi-million-line CATIA C++/Linux codebase, and earlier a Data Acquisition Engineer at Renault. Those years shaped how I write research code today.