Freelancer profile translated to English.

Description

I have been solving image and signal processing problems for over ten years. I worked in R&D for medical imaging, in the nuclear sector, and in photography — after a PhD in astrophysics obtained in 2020. I design low-level solutions, rooted in measurement physics and detection theory, where noise and artifacts cause generic methods to fail.

Industry field of expertise

Languages

French
Native or bilingual
English
Native or bilingual

Workplace preferences

Can work on-site

Paris (up to 50km)

Indépendant
Consulting & Development
TECH
May 2026 - Today (3 months)
Paris, France
Automatic cropping of scanned films — computer vision product in production: robust detection (perforations, light leaks, tilted frames, underexposure) and film-wide consensus. Python; in production for several months — approximately 12 films processed per day.
Computer Vision Signal Processing Software Architecture Python OpenCV
Laboratoire argentique
Co-founder
FILM AND AV
May 2025 - April 2026 (11 months)
Paris, France
Design and operation of a complete workflow, from shooting to file:

- Design and construction from scratch of a complete analog lab using consumer equipment (€4,000) — capacity ~50 35mm films/day over 24 m2.
- Scanner and end-to-end processing chain; in-house densitometric tracking software: measurement of Dmin and Dmax on reference test strips.
- Manual monitoring and replenishment of development chemicals
Computer Vision OpenCV
Médiane Système
R&D Engineer (Consultant)
MEDICAL
January 2022 - August 2024 (2 years and 7 months)
Grenoble, France
Medical Imaging — Signal Processing:
- Electromagnetic correction of the imaging chain for large-format X-ray detectors: suppression of sensor-coupled artifacts (mains and harmonics, motors, RF, screens), without degrading clinical information.
- Performance evaluation of the chain: clinical information preservation metrics, artifact suppression quantification, non-regression tests on representative image sets.
- 100% deterministic pipelines with bounded execution time — under one second, with no perceptible delay for the operator.
- Algorithms dedicated to each defect class (periodic with harmonics, transient spikes, block periodicity), arbitrated by a decision tree.
- Hardware-level detection: exploitation of the sensor's readout order to refine noise estimation.
- Specifications: translation of algorithms into a detailed low-level implementation plan, understandable by C/C++ developers and non-signal processing specialist system engineers.

Nuclear Instrumentation — R&D Valorization:
- Interface between technical experts (design and qualification of safety and protection systems — neutron flux monitoring) and public funders.
- Translation of R&D strategy into quantified ROI: conversion of technical levers (code and test generation, quality control, tools avoiding research reactor campaigns) into financial stakes, in line with French nuclear strategy.
- Drafting of valorization dossiers and technical summaries for non-specialist decision-makers.
Algorithms Medical Imaging Signal Processing Image Processing Python