Wissam Boussella

I am a machine learning and software engineer focused on training, deploying, and optimizing deep learning models for constrained systems. My background spans signal processing, 3D rendering/reconstruction, and low-level systems integration.

At CEA I contributed to an open-source core project in advanced C++, exported complex ONNX models with custom nodes, and developed ultra-low precision C/C++ kernels for 4-bit quantized networks up to 60M parameters. I also generated compilable C via Python/Jinja2 and validated millisecond-level inference on STM32 hardware.

Earlier, I enhanced 2D CAD tooling by integrating Bezier curve support into a simulation product, implementing robust backend geometry handling and GUI integration. I also completed research work applying self-supervised learning techniques with OpenCV and PyTorch.

I build reproducible, test-driven systems—authoring 50+ unit tests and Docker CI/CD—and have delivered embedded deployments from training and quantization to automated ONNX-to-C conversion and hardware optimization.