My work experiences

Lead developer and architect for the Aidge Eclipse platform (successor to N2D2), a comprehensive framework for optimizing and deploying Deep Neural Networks (DNN) on constrained embedded targets.

• Framework Architecture: Orchestrated the full lifecycle of the Aidge/N2D2 ecosystem, including the development of core engines for quantization, code generation, and cross-platform model deployment.
• High-Performance Computing: Developed custom high-performance kernels and operators in C++ and CUDA/cuDNN for object detection and semantic segmentation, specifically optimized for autonomous driving time constraints.
• Embedded Deployment: Optimized and deployed models on diverse hardware including NVIDIA Jetson Orin (via TensorRT), STM32, and specialized neural accelerators (NeuroCorgi, PNeuro).
• Developer Ecosystem: Built a comprehensive Python SDK using PyBind11, enabling seamless interoperability between C++ backends and the research ecosystem (PyTorch, ONNX, NumPy).
• Engineering: Modernized the development workflow by implementing GitLab CI/CD pipelines and Docker environments, significantly improving framework stability and release velocity.
• Technical Mentorship: Directing the research of PhD and Master's students on cutting-edge topics including LLM quantization (LLaMA) and Vision Transformer optimization.

Final year research internship focused on optimizing the generation of linguistic summaries from massive databases using fuzzy logic and advanced data structures.

• Algorithmic Optimization: Addressed the exponential complexity of generating conjunctive fuzzy properties (e.g., "Most crows are black AND tall") by moving away from linear database scans.
• Research & Implementation: Conducted a bibliographic study on multidimensional data representation and implemented three complex histogram methods (MHIST, GENHIST, and STHOLES) in Python to mitigate the "curse of dimensionality."
• Infrastructure & Reproducibility: Designed and deployed a dockerized REST API using Flask to wrap the research algorithms, ensuring portability and reproducibility of the results. team.

International research internship at the Intelligent Electronics and System-On-Chip (IESOC) laboratory, focusing on the intersection of biomedical signal processing and edge computing.

• Signal Processing: Developed a Python-based pipeline to process raw ECG (Electrocardiogram) and PPG (Photoplethysmogram) signals, implementing peak detection algorithms to extract robust features for affective computing.
• Deep Learning Development: Designed and trained a Convolutional Neural Network (CNN) using Keras to classify human emotional states based on physiological responses.
• Edge AI Deployment: Successfully ported and optimized the trained model for real-time inference on the NVIDIA Jetson Nano.

• Systems Supervision: Contributed to the upgrade of network supervision services using Centreon during a major ISD merger.
• Automation: Developed PowerShell scripts to audit and manage the Active Directory, identifying and removing inactive user accounts.
• Responsibility: Managed critical infrastructure tasks where high precision was required to maintain service continuity for workers.