Freelancer profile translated to English.
Back to original languageAbout Philippe
Looking for a specialist capable of designing, training, and optimizing Computer Vision or medical AI models for your products or R&D projects? I will support you from start to finish: from needs analysis to an operational, robust, and documented model.
🚀 How I can help you
- Implement advanced learning strategies: Vision Transformers, self-supervised learning, multimodal learning, contrastive learning.
- Optimize your pipelines: GPU acceleration, CUDA, C++ optimization if necessary.
- Create internal tools: annotation interfaces, analysis dashboards, automatic dataset cleaning.
💡 My added value
- Expertise in medical AI (MRI, mammography, ultrasound, CT), acquired at Siemens Healthineers, Lunit, GE Healthcare, and DentalMonitoring.
- A structured R&D approach: reproducibility, scientific rigor, clear documentation.
- A strong ability to combine research and application.
- Experience in scientific publication (first author MICCAI) and Deep Learning teaching.
🏆 Examples of projects completed
- Brain tumor segmentation: multi-label segmentation pipeline (DICE 93%), scientific publication.
- Breast cancer risk prediction: improvement of +2 to +3% AUC (Lunit).
- Cardiac segmentation: multimodal ViT pre-training (CT/MRI/UltraSound) and +4% Dice gain (Siemens).
- Embedded model for dental hygiene assessment on mobile (DentalMonitoring).
- GPU optimization (CUDA): reduction of processing time by up to 97% on an object detection pipeline.
🧪 Types of projects I undertake
- Development and optimization of models in PyTorch or TensorFlow
- Computer vision pipelines (medical or non-medical)
- Dataset preparation, annotation error detection
- Deployment of experimental or production-ready models
- High-performance implementations (GPU, CUDA, C++)
If you are interested, do not hesitate to contact me!
French
Native or bilingual
English
Native or bilingual
Can work on-site
Paris (up to 10km)
Experience
- Laboratoire de Recherche de l'EPITA (LRE)Research InternMEDICALMay 2022 - February 2023 (9 months)Paris, FranceProposed the use of morphological max-trees for medical annotation tasks, in collaboration with GE HealthCare.Calculated shape attributes for max-tree nodes to aid in detecting relevant structures.Developed a Python web interface for an interactive MRI annotation task.Achieved an average DICE score of 83.17% in a cardiac MRI annotation task.Published a research paper proposing an innovative solution for brain tumor segmentation.
- DentalMonitoringMachine Learning EngineerMEDICALJuly 2025 - Today (11 months)Paris, FranceDeveloped a high-performance classification network for dental hygiene assessment on mobile devices using TensorFlow.Conducted performance tests of orthodontic models on patients in real clinical conditions.Reorganized an old database into a new standardized format to optimize model training.
- Lunit Inc.AI Research EngineerMEDICALAugust 2024 - February 2025 (6 months)Séoul, South KoreaLed research on breast cancer risk prediction in mammograms, utilizing contrastive learning to enhance accuracy in PyTorch.Eliminated the need for pre-training by aligning high-risk unlabeled embeddings with labeled embeddings during training.Improved feature representations, resulting in a 2% to 3% increase in AUC, demonstrating enhanced model performance.Conducted in-depth analysis of the mammography dataset and provided actionable insights for precise decision-making to optimize model performance.
Recommendations
Be the first to recommend Philippe
Help this freelancer shine by sharing your experience working together.
These freelancer profiles also match your criteria
Agatha Frydrych
Backend Java Software Engineer
4.7
(3)
2
Baptiste Duhen
Fullstack developer
4.6
(4)
5
Amed Hamou
Senior Lead Developer
4
(2)
7
Audrey Champion
Web developer
4.3
(3)
4
Education
- Engineering DegreeEPITA2023Informatique avec specialization deep learning
- Introductory Workshop to Deep Learning [Workshop Material]Google DeveloperIntroductory Workshop to Deep Learning [Workshop Material]