Principal Research Scientist · Paris, France
I develop large-scale generative and multimodal systems for image editing and generation, building on diffusion and transformer-based foundation models. My work spans problem formulation, model design, large-scale training, rigorous evaluation for production deployment.
Over the past few years, I have developed and shipped generative AI systems for image editing and generation that are used by millions of users and have helped secure contracts worth seven figures in annual revenue.
I currently lead the Image Research team at Jasper, developing state-of-the-art generative and multimodal models. Previously, I was a Research Scientist at Stability AI. I hold a PhD in Computer Vision and Machine Learning from Inria.
Interests Generative image editing · Diffusion models · Multimodal learning · Distributed training
Research work that I directly contributed to, spanning research, code, method development, and training at scale.
An outpainting system that extends images beyond their borders with coherent results. Built on and fine-tuned from Flux to improve generation quality, reduce hallucinations, and better match user preferences with a distilled, lower-latency variant for faster inference.
Proposed a method for controllable shadow generation in product photography, enabling users to control shadow direction, softness, and intensity. I also created a public benchmark from scratch to evaluate the quality and controllability of generated shadows.
Direct application of our Latent Bridge Matching approach (ICCV 2025) for generative object removal, producing high-quality results in near real time. Extensively fine-tuned on large-scale datasets for robust, production-ready performance.
Built a rendering pipeline from scratch on top of Blender to generate perfectly annotated, large-scale datasets for training multimodal generative models. It powers the training of several of our models, including object removal, relighting, and shadow and reflection prediction.
Trained a custom SDXL model allowing the users to replace the face of a source image with the face of a target image.
Generative relighting of images and objects in real time, with interactive control over the number of lights and their properties, including color, intensity, and position. This tool went viral on various social media platforms with millions of views and likes.
Latent Bridge Matching for Image-to-Image Translation
ICCV 2025 (Spotlight) · second author
Flash Diffusion: Accelerating Diffusion Models
AAAI 2025 (Oral) · second author
Controllable Shadow Generation with Single-Step Diffusion Models from Synthetic Data
arXiv 2024 · first author
MONET: an open dataset of 100M+ text–image pairs
NeurIPS 2026 · third author
Jasper · France / USA
Technical lead of the image research team, focused on developing and deploying generative and multimodal models for image editing and generation.
Jasper · France / USA
Owned the end-to-end development of multimodal, transformer, and diffusion-based image editing systems, from model design and training to evaluation and production, powering features used by millions of users.
Stability AI · France / UK
Collaborated with researchers from the original Stable Diffusion team to develop creative applications based on adapted Stable Diffusion models.
Clipdrop · France
Developed and contributed to a suite of production image editing systems used by millions of users.
Inria · France
Developed object detection models trained entirely on synthetic data from game-engine rendering and simulation pipelines, improving robustness under domain shift. The work supported an Inria-incubated startup that went on to raise €2.5M in seed funding.
Inria · France
Computer vision, machine learning, and computational geometry for map generation from satellite imagery.
Feel free to reach out to me via the links below.