CV

AI Scientist - Algorithm Developer

LeddarTech

Developing simulation-driven algorithms for Neural Radiance Fields (NeRF) and 3D Gaussian Splatting (3DGS).

• Built data pipelines for NeRF and 3DGS training and rendering.
• Developed 3D scene reconstruction tools and transformation algorithms.

Computer Vision & AI Scientist

INO (National Optics Institute)

Led computer vision initiatives including segmentation systems and annotation tools. Delivered high-impact AI solutions for industrial applications.

• Developed custom instance segmentation models with PyTorch, improving model precision and runtime.
• Implemented and deployed a robust TensorFlow-based segmentation pipeline with 20% IoU improvement.
• Led deployment and customization of CVAT on AWS with semi-automatic annotation features, reducing labeling time by 40%.
• Engineered reusable CV toolkits and documentation adopted by multiple internal teams.
• Achieved 98% model accuracy through advanced data augmentation and optimization.

Data Scientist

BI Expertise

Delivered cloud-based image classification solutions on Google Cloud Platform.

• Designed and tested classification models using TensorFlow on GCP.
• Improved accuracy by 10% through systematic fine-tuning.
• Boosted business insights and customer satisfaction through automated workflows.

Bachelor

University of Shanghai for Science and Technology

Thermal Energy and Power Engineering

Master

University of Rouen

Energy, Fluids and Environment

Master

University of Rouen

Laser Diagnostics and Optical Metrology

PhD

Université Laval

Electrical Engineering

• Infrared Vision for Non-Destructive Testing

Multi-Sensor Fusion and 3D Scene Reconstruction

Developed and deployed sensor fusion algorithms and 3D reconstruction systems for automotive perception.

• Integrated multi-sensor inputs (e.g., LiDAR, RGB) for real-time 3D scene reconstruction.
• Designed flexible transformation pipelines for heterogeneous sensor datasets.
• Supported R&D simulation with reproducible 3D environments.

Simulation Framework with NeRF & 3D Gaussian Splatting

Built a data-driven simulation framework for autonomous driving using Neural Radiance Fields (NeRF) and 3D Gaussian Splatting techniques.

• Developed pipelines for data ingestion, transformation, and augmentation tailored for autonomous vehicle simulation.
• Integrated NeRF and 3DGS models with custom dataloaders and sensor metadata.
• Translated cutting-edge academic methods into production-grade systems.

Instance Segmentation with Custom Model Heads (PyTorch)

Led development of a state-of-the-art instance segmentation system with custom model heads, optimized for industrial vision tasks.

• Achieved 98% accuracy through architecture tuning and data augmentation.
• Improved inference performance and generalizability across datasets.
• Packaged reusable modules for future vision projects at INO.

CVAT on AWS with Semi-Automatic Labeling

Engineered and deployed a customized CVAT platform on AWS to support internal data annotation workflows.

• Reduced labeling time by 40% through semi-automatic annotation integration.
• Customized CVAT backend and UI for domain-specific needs.
• Provided documentation and training adopted across INO departments.

TensorFlow-based Semantic Segmentation Pipeline

Designed and deployed an image segmentation system with TensorFlow, improving segmentation quality and efficiency.

• Enhanced IoU metrics by over 20% using improved preprocessing and architecture.
• Deployed pipeline for batch processing across multiple projects.
• Automated model evaluation and retraining procedures.