Description

Machine Learning Engineer with extensive experience in computer vision and, more recently, in the fragrance industry. Recently promoted to Head of AI means leading a team, creating tech strategy, inter-team coordination, and presenting strategic results to stakeholders while continuing to build production-ready machine learning solutions. Transformed company strategy from mostly heuristic-based to learning-based. Improved performance in all significant metrics - perceptual similarity, perceptual additivity, reformulation quality, etc.

Industry field of expertise

Languages

English
Native or bilingual
Russian
Native or bilingual
French
Native or bilingual
Spanish
Conversational

Workplace preferences

Can work on-site

Aix-en-Provence (up to 50km)

Moodify
Head of AI
CHEMICAL
September 2025 - Today (10 months)
Paris, France
• • Replaced predominantly heuristic, non-learning pipelines with learning-based approaches, with a strong emphasis on data quality and data integrity.
• • Developed a scent similarity model through ensemble learning over non-linear models, achieving substantial gains in correlation and MAE.
• • Supervised the development of a learned scent-perception additivity model, significantly reducing MAE compared to prior heuristic approaches.
• • Supervised the design and development of a production database, focusing on data integrity.
Machine learning Management d'équipe Gestion de projet Développement Python
Moodify
Senior AI research scientist
CHEMICAL
January 2025 - September 2025 (8 months)
Paris, France
• • Trained, evaluated and tested probabilistic models for scent similarity prediction using perceptual data.
Machine learning Développement Python
Meero
Computer Vision Engineer
TECH
January 2019 - January 2025 (6 years)
Paris, France
• • Accepted speaker at GenAI Days 2025 on behalf of Meero, presenting the Magic Eraser feature and its underlying generative modeling pipeline.
• • Led research and development of the Magic Eraser feature, training a custom latent diffusion in painting model with strong focus on quality and runtime (model compilation, heuristic design, and custom distillation). Became the app's #1 feature and ultimately became the key driver in the acqui sition of Meero's ProperShot team by NodalView.
• • Led development of the Home Staging feature, training custom latent diffusion models to automat ically furnish interior scenes.
• • Led development of the Wall Painting feature, training a custom segmentation model for wall de tection and applying a pre-trained diffusion model for realistic recoloring.
• • Collaborated on a CIFRE PhD project involving private-data image editing, internship supervision on inpainting, and technical consulting.
• • Developed custom super-resolution and artifact removal models based on ESRGAN.
• • Implemented noise removal and sharpening using an FMLE-based algorithm, parallelized on CPU and GPU with CUDA.
Machine learning Développement Python Speaker