Papers
3
Total Citations
20
H-Index
2
About
Angeliki Kritikakou is a leading researcher at the intersection of reliable AI and safety-critical embedded systems. Her work focuses on ensuring that Deep Neural Networks (DNNs) and Deep Reinforcement Learning (DRL) policies remain trustworthy when deployed on resource-constrained hardware, particularly in autonomous robotics, aerospace, and healthcare. She pioneered **harDNNing**, a machine-learning-driven framework for fault tolerance assessment and protection of DNNs (12 citations), which systematically evaluates and hardens neural networks against hardware faults. Her investigations into the **neutron sensitivity of DRL policies on EdgeAI accelerators** (6 citations) revealed critical vulnerabilities in Google’s Coral Edge TPU, demonstrating that even state-of-the-art edge devices can produce catastrophic errors under radiation. More recently, she has developed methodologies for the **reliability assessment of large DNN models** that intelligently trade off performance for accuracy (2 citations), enabling practical deployment in real-world autonomous systems. Kritikakou’s contributions are vital for bridging the gap between high-performance AI and the rigorous dependability demands of safety-critical applications, making her a key voice in the emerging field of resilient edge intelligence.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3