Freelancer profile translated to English.

Description

🎓Charlie is...

A Machine Learning Engineer specializing in Computer Vision — object detection and 3D vision — and MLOps, for models that truly perform in production.
A graduate of EPITA, a renowned engineering school in AI and Computer Vision (MSc Data Science, ML & Computer Vision).
And who assists companies in the design, industrialization, and monitoring of vision models.

✨ Building your vision pipeline, from idea to production

Need to detect and locate objects in your images/videos, reconstruct a 3D scene, or get a vision model into production without exploding your inference costs?

ML Engineer specializing in Computer Vision, I design, train, and deploy measurable, production-ready object detection and 3D vision pipelines: API/serving, clear metrics, operational documentation, handover to your team.

Use cases: object detection/counting (security, industry, retail), 3D reconstruction and calibration, industrializing a model stuck in a notebook (serverless, ONNX/TensorRT, monitoring).

🔄 A method based on continuous improvement

A model is never "finished" at deployment: I measure before optimizing, I monitor drift in production, I reimplement critical components rather than stacking libraries to be able to fix what matters, and I document every decision so that your team remains autonomous.

✅ Working with me means relying on:

Comprehensive support (data collection, training, deployment, monitoring)
Proven expertise in object detection and 3D vision
Robust production infrastructure
Explained monitoring: I help you understand each step for total control over the strengths and weaknesses of your project.

✉️ Send me your project and I'll propose a mini action plan to achieve your goals!

Industry field of expertise

Languages

French
Native or bilingual
English
Fluent

Workplace preferences

Remote only

Primarily works remotely

Yield Studio
Machine Learning Engineer
SOFTWARE PUBLISHING
October 2025 - November 2025 (1 month)
Paris, France
Development of an AI system for detecting airplanes on billboards
Methodical exploration of 4 classical computer vision approaches (HSV, ORB template matching, contour detection) to establish a baseline with 65-70% accuracy.
Construction of a complete PyTorch training pipeline with data augmentation (rotation, scale, brightness) and transfer learning on MobileNetV2.
2-phase fine-tuning: freezing pre-trained layers, then full training with adaptive learning rate.
Model optimization and deployment (PyTorch → ONNX) for WebAssembly execution in the browser, final size 8.47 MB, latency 50-100ms.
Development of a complete web application (JavaScript, HTML) with batch testing interface and real-time metrics (confusion matrix, precision, recall, F1-score).
Performance improvement of +20% compared to the classical baseline: accuracy 72.4%, precision 100%, recall 63.6%.
Exhaustive technical documentation and recommendations for scaling up to production (target 90%+ accuracy).
Computer Vision Python Pytorch Machine Learning OpenCV
WearIT
Co-Founder & Machine Learning Engineer
FASHION AND COSMETICS
January 2024 - Today (2 years and 7 months)
- Development of a frontend and backend POC for the solution (React, NestJS, MongoDB, AWS EC2).
- R&D on 3D Virtual Try On using the OCCNet model (PyTorch).
- Construction of a dataset of 3 million images (mannequin-clothing pairs) via web scraping (JS, MongoDB, AWS S3).
- Automation of labeling (type classification, texture, color, morphology) using PyTorch models.
- Multi-GPU, multi-node training (32x A100) for the creation of two proprietary AI diffusion models (PyTorch): SSIM 0.8170, PSNR 15.57, FID 4.88.
- Production deployment of models with ONNX/TensorRT on AWS EC2.
- Development of an application
Pytorch Python Generative AI Deep Learning Machine Learning
Lumyosa
Data Scientist & Software Engineer
September 2022 - February 2023 (5 months)
- Design of a voice gender classification model (MFCC + boosting tree, Scikit-Learn), F1 score: 98%.
- State-of-the-art study on vocal and textual sentiment analysis.
- Development of a CNN for emotion classification (TensorFlow), F1 score: 60% (data drift).
- Creation of a customer feedback analysis platform (Angular, PHP) integrating AI models for sentiment analysis.
TensorFlow Deep Learning Data Science Machine Learning Python