Abstract: AI at the space edge fails across layers: radiation-induced faults, aging, thermal stress, and silent data corruption can compromise processors and accelerators, while learned models may still produce plausible but unsafe outputs. This talk argues that resilience should be treated as a service plane rather than as a fixed hardware feature or a software patch. Drawing from ongoing work on reconfigurable FPGA/SoC accelerators, post-quantum avionics services, hardware security, entropy generation, and NASA JPL’s F′ flight-software framework, the talk presents a cross-layer approach in which hardware capabilities are exposed as commandable, telemetry-visible services. At the hardware layer, runtime-selectable accelerator modes and mappings allow a system to trade performance, redundancy, and containment as mission conditions change. At the software layer, telemetry-grounded assurance provides the evidence needed to gate safety-critical autonomous decisions. The central position is that software cannot truly fix bad hardware, but cross-layer systems can detect, bound, isolate, and route around hardware-induced failures before they become mission failures.

Dr. Mohamed El-Hadedy is an Associate Professor of Electrical and Computer Engineering at California State Polytechnic University, Pomona, and the founding director of the Reconfigurable Space Computing Lab (RSCL). His research lies at the intersection of autonomic systems, UAV reliability, post-quantum cryptography, and adaptive hardware/software co-design for space and defense applications.

Dr. El-Hadedy leads multiple federally funded projects, including a U.S. Navy grant on accelerating NLP in drone swarms using Kubernetes-based orchestration, a DOD initiative on energy-efficient secure computing with FPGA clusters, and an AFRL-sponsored effort to create a new generation of flight controllers infused with edge AI and lightweight cryptography.

His lab has been a multi-year finalist in NASA MINDS, with teams pioneering CubeSat drones, ionic thrusters, and Tesla coil-enabled UAV systems. Dr. El-Hadedy’s recent research introduces runtime thermal and vibration-based diagnostics for autonomous aerial platforms, integrating LLMs, YOLO, and cryptographic primitives in real-time edge deployments.

He holds multiple U.S. patents, several best paper awards (SoCC, IEEE conferences), and is passionate about building systems that adapt, heal, and secure themselves at the edge — from Earth to orbit.