At dAIEDGE, we continue our Meet the Winners series by showcasing the people and projects addressing some of today’s most pressing challenges through AI at the Edge. In this second interview, we feature Sotirios Athanasoulias, one of the winners of our First Open Call, and the developer of the project ENERGIZE.
As the global energy crisis intensifies, finding practical and scalable ways to reduce energy consumption has become a priority. ENERGIZE directly addresses this challenge by helping users better understand how energy is consumed at home. The project focuses on Non-Intrusive Load Monitoring (NILM)—a technique that disaggregates total household energy consumption into appliance-level insights using only data from a single smart meter, without the need for additional sensors.
What makes ENERGIZE particularly innovative is its approach: instead of relying on cloud-based processing, it enables decentralized, edge-based energy disaggregation. By leveraging computationally efficient, edge-optimized AI models, the project delivers accurate, low-latency insights directly on-device, enhancing privacy while reducing infrastructure costs.
One of the key technical challenges lies in adapting state-of-the-art deep learning models, typically resource-intensive, to run on low-cost, resource-constrained hardware. To overcome this, ENERGIZE explores advanced model optimization strategies, including compression techniques and quantization, significantly reducing memory footprint and computational requirements while maintaining reliable performance. This enables real-time feedback to be generated locally, making the solution both scalable and accessible.
During his journey within dAIEDGE, Sotirios developed his project at Foundation for Research and Technology – Hellas. We would like to extend our sincere thanks to FORTH for their role as a hosting institution, providing essential support and technical guidance throughout the programme.
This collaboration played a crucial role in shaping the direction of ENERGIZE, from refining the initial concept to defining realistic deployment conditions. In particular, identifying target hardware constraints—such as strict memory and latency requirements—highlighted the need for aggressive model compression and directly influenced the project’s technical approach.
ENERGIZE exemplifies the type of innovation that dAIEDGE aims to support: solutions that combine cutting-edge AI with real-world applicability, addressing societal challenges such as energy efficiency, sustainability and privacy.
This interview is part of our ongoing series highlighting the winners of our Open Calls. Stay tuned for more insights and stories on our YouTube channel.