Jules Perret

I am a PhD researcher and research engineer in gravitational-wave astrophysics at the Astroparticle and Cosmology Laboratory (APC, CNRS). My work revolves around one central problem: how do we extract reliable astrophysical information from gravitational-wave signals, efficiently enough to keep pace with current and future detectors?

My current work is focused on LISA, the ESA-NASA space interferometer scheduled for launch in the early 2030s. LISA will observe massive binary black holes across cosmic history, but its data analysis pipeline is a genuine computational challenge. I work on parameter estimation for these systems, both on the sampling side and, increasingly, on building fast, differentiable likelihoods that run efficiently on CPU and GPU using JAX.

Research

My work sits at the intersection of Bayesian statistics, scientific computing, and gravitational-wave physics. The central bottleneck in gravitational-wave parameter estimation is the likelihood: evaluating it is expensive, and you need it millions of times. My current focus is on developing fast likelihood implementations for massive binary black holes in the LISA band, written in JAX to exploit autodifferentiation and run seamlessly on CPU and GPU.

On the inference side, I design and implement sampling algorithms (HMC variants and parallel tempering schemes) that can handle the multimodal, high-dimensional posteriors LISA will produce. During my PhD, I also worked on neural network-based approaches to accelerate sampling for ground-based detector sources.

Projects

Project Description Stack
JaxMBHB (current) Fast time-domain LISA likelihood for massive binary black holes JAX, XLA, TDI
PT-HMC Parallel tempering HMC for multimodal posteriors on LISA sources JAX, GPU
DeepHMC (PhD) Neural metric learning to accelerate HMC for GW parameter estimation PyTorch, CUDA
GW Event Visualizer Interactive GWOSC catalog dashboard, auto-updated via CI Python, GitHub Actions
Bayesian Inference Guide Interactive intro to Bayesian methods and sampling algorithms MCMC, HMC, Nested Sampling

Beyond the Lab

I am a member of the Société Astronomique de Bourgogne, where I contribute to public outreach and astronomy education. Explaining orbital mechanics to a curious twelve-year-old is a surprisingly good test of whether you actually understand it.

Outside physics, I build things: drones, UAVs, whatever requires soldering and patience. It is a useful counterweight to work that lives entirely in abstract probability spaces.

Contact