Freelancer profile translated to English.
Back to original languageAbout Maël
Expert Engineer in Artificial Intelligence & Computer Vision | Deep Learning | Machine Learning
EPITA Engineer, I combine technical expertise and international experience (Bosch Singapore, Huawei) to develop your AI and computer vision solutions.
Technical expertise:
- Artificial Intelligence & Deep Learning: TensorFlow, PyTorch, scikit-learn
- Computer Vision: OpenCV, image and video processing
- Edge Computing & Embedded Systems: Optimization, edge deployment
- Data Science: numpy, pandas, data analysis and valorization
Significant achievements:
- Improvement of embedded vision system performance from 70% to 94% (Bosch)
- Optimization of AI models for constrained systems: 2s -> 0.01s prediction time
- Development of complete image processing pipelines
- Contribution to several company-validated invention reports
Services:
- 🔍 Design and development of AI/Vision solutions
- 📊 Proof of Concept and rapid prototyping
- 💻 Model optimization for embedded systems
- 📈 Technical audit and consulting
- 🎯 Training and support
Technologies: Python, C++, CUDA, TensorFlow, PyTorch, OpenCV, scikit-learn, numpy, pandas, PHP
French
Native or bilingual
English
Native or bilingual
Spanish
Conversational
Can work on-site
Paris (up to 30km)
Experience
- Joris LIMONIER
On Malt
AUTOMATIC DETECTION OF JUDO SCORES BY DEEP LEARNINGSPORTSOctober 2025 - October 2025MISSIONDevelop an MVP for automatic detection of scoring moments in judo fight videos to reduce manual viewing time by 80%.RESPONSIBILITIES & ACTIONS- Design end-to-end ML pipeline: IJF API scraping → real-time detection
- Conv1D + LSTM architecture for video sequences (225 MediaPipe pose + hand features)
- Feature engineering with hip-centered normalization and 2s time windows
- 4 detection modes: single, batch, streaming with circular buffer
- Interactive CLI orchestrating 6 modules (scraping, annotation, training, detection)
- Resolution of critical blockers: Docker → pyenv+venv, Keras 3.11→3.8.0
- Time extraction optimization: 2h → 60 min with MediaPipe 10fps
- Scrape IJF API with diverse selection (50 fights, automatic annotations)
- OpenCV manual annotation GUI tool for timestamp recalibration
- Automatic generation of no-scores with empirically optimized 10:1 ratio
- Intelligent feature cache system (avoids re-extraction, 10x gain)
RESULTS- Specificity 81.85% (target 70% exceeded by +11.85 points)
- Accuracy 78.66% on dataset of 49 videos, 53 scores, 567 no-scores
- Reduction of ~80% manual viewing time (532/650 moments correctly filtered)
- Complete functional and documented pipeline (README, changelog, CLI guide)
- 8 modular production-ready Python scripts (~2500 lines)
- Complete evaluation system: TP/FP/FN/TN confusion matrix, 5 metrics
- SKILLS
- Deep Learning (Conv1D, LSTM, feature engineering, dataset balancing)
- Computer Vision (MediaPipe pose estimation, real-time video processing, spatial normalization)
- Software Engineering (modular architecture, interactive CLI, caching, documentation)
- Data Engineering (REST API scraping, ETL pipeline, synthetic data generation)
- Problem Solving (comparative technology analysis, blocker resolution, informed trade-offs)
- Agile methodology (rapid iterations, empirical testing, data-driven decisions)
- Robert Bosch ASEANAI / Electronics InternTECHFebruary 2024 - August 2024 (6 months)Singapour, Singapore
- Development and optimization of AI models for embedded home automation systems, improving accuracy from 70% to 94%
- Improvement of model inference times from 2 seconds to less than 0.01 seconds
- Implementation of a hybrid Deep Learning/Machine Learning approach to optimize performance on Raspberry Pi
- Complete management of the data pipeline: collection, preparation, and analysis of training datasets
- Proposal and implementation of a strategic migration to traditional ML models to improve efficiency on constrained systems
- Huawei Paris Research centerResearch internTECHSeptember 2022 - January 2023 (5 months)Boulogne-Billancourt, France
- Development of monitoring and debugging tools for real-time operating systems (RTOS)
- Analysis and optimization of kernel performance for critical embedded systems
- Creation of an analysis tool suite to improve the reliability and traceability of RTOS in development
- Participation in improving debugging processes for system development teams
Reviews
Joris
Très bien - Joris LIMONIER
Reviewed on 12/3/2025
Good performance on an Artificial Intelligence and computer vision project.
Recommendations
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 DegreeEPITA2024Diplôme d'Ingénieur en Informatique - EPITA Spécialisation Intelligence Artificielle & Imagerie Numérique | 2024 Formation d'excellence en ingénierie informatique Majeure : IMAGE (Intelligence Artificielle et Vision par Ordinateur) Domaines d'expertise acquis : - Deep Learning et Machine Learning - Traitement d'images et Computer Vision - Développement logiciel avancé - Systèmes embarqués et optimisation CTI (Commission des Titres d'Ingénieur) - Grade Master