After KAUST Academy
Mohammed works at SAAI in Riyadh and represented it at LEAP 2026. Mohammed completed the Samsung Innovation Campus AI program with Misk Skills in September 2026, earned the CAIE certification, and judged the BUILDx Hackathon in October 2026.
Summer project
Artificial Intelligence Specialization · Cohort 2025
Propagated Gradient Based Learnable Dynamic Compression and Energy Optimisation for Edge AI on Microcontrollers
ML models demand substantial computation, memory, and energy resources, hindering their deployment on edge devices with limited hardware. To tackle these constraints, we introduce a state-of-the-art technique, “S More Than You Need,” a propagated-gradient-based learnable dynamic compression method that obtains a subnetwork that performs as well as the full network, reducing size by 35× to 88× without loss of accuracy. Our method guides the gradient toward convergence, providing optimal accuracy and compact size. We reinforce our work with long-term solutions such as motion sensors to extend battery life and reduce computational load, achieving up to 90% energy savings, and lightweight communication protocols such as MQTT and LoRa. Additionally, we assess our methods on several microcontrollers, providing reliable performance across models such as LeNet, ResNet, MobileNet, and LightNN.
Mentor: Rytis Paškauskas
With KAUST Academy
- Cohort 2025
Artificial Intelligence Specialization
International Centre for Theoretical Physics · Artificial Intelligence Summer Internship
Want your entry updated or removed? Contact us.