Bringing intelligence closer to real-world interaction often means dealing with noise, uncertainty, and strict hardware limits all at once.
That is exactly the space explored by NAVIR, the project developed by Mike Karamousadakis as part of the Second Open Call of dAIEDGE.
NAVIR focuses on neuromorphic computing at the Edge, using low-power, real-time multimodal spiking neural networks (SNNs) to enable more robust human–robot interaction in challenging and noisy environments. The goal is not just performance, but resilience under real operating conditions where traditional models often struggle.
In this context, efficiency is not optional, it is a requirement. The project explores how neuromorphic approaches can reduce power consumption while maintaining responsiveness, making them suitable for deployment in embedded and edge systems.
During the programme, Mike worked in close collaboration with Thales, whose mentorship and technical guidance were instrumental in shaping the implementation of the project. We would like to sincerely thank Thales for their role as hosting institution and for their continuous support throughout the process.
Alongside this, access to the dAIEDGE-VLab provided an additional layer of validation, allowing experiments to be tested under realistic computational constraints and supporting the transition from conceptual design to practical evaluation.
NAVIR reflects a key direction of dAIEDGE: enabling intelligent systems that are not only powerful, but also efficient, adaptive, and deployable in real-world edge scenarios.