Issam Sayyaf

I build deep-learning models for noisy, multi-modal sensor time series. As a Senior Algorithm Engineer on the Edge AI team at TDK InvenSense in Grenoble, I train CNNs, RNNs/LSTMs and Transformers for forecasting, anomaly detection, gesture recognition and spatial tracking — then get them running on-device, quantized and pruned, through ONNX and TensorFlow Lite.

My PhD (Université Gustave Eiffel, 2026) was on physics-informed anomaly detection for navigation time series, and I surveyed the field for IEEE Sensors Journal in 2025. Before that I spent several years building production Yocto systems and Linux kernel drivers — which is why the models I ship tend to survive contact with real hardware.

I write about this occasionally. The fastest way to reach me is email.

Recent writing