Think about listening to a person speaking to you in a foreign language. Would you be interested in the word-for-word translation or the meaning of the sentence? I would focus on understanding the meaning and I bet you would do the same. Semantic communications apply this concept to wireless networks. Instead of reconstructing data bit-by-bit as in traditional networks, we are interested in obtaining the most important features of the input according to a specific objective. In this talk, I will present our approach, named PhyDNNs, to perform a downstream AI task at a receiver using the features extracted by a transmitter. Task-relevant information is extracted through a neural network which is executed at the physical layer, i.e., it directly provides the waveform to be transmitted. Robustness to channel noise is achieved by integrating the channel into the end-to-end training of the transmitter-receiver architecture. Experimental results show that PhyDNNs can reduce the end-to-end inference latency and power consumption by up to 48× and 13× respectively, while keeping the accuracy within 7% of the state-of-the-art approaches. Building on these results, I will present our Semantic Multiplexing approach to increase the scalability of the system. Our strategy allows processing multiple task inputs simultaneously and combining the relevant information into a single semantic waveform that can be transmitted through single- or multi-antenna systems. Our experiments indicate that semantic multiplexing reduces latency, energy consumption, and communication load by up to 8×, 25×, and 54×, respectively, compared to existing baselines while maintaining comparable performance.
Francesca Meneghello is a Principal Research Scientist at the Institute for Intelligent Networked Systems (INSI) at Northeastern University, USA. She worked as an Assistant Professor at the Department of Information Engineering at the University of Padova, Italy. Her work focuses on wireless communications and machine learning, contributing to defining and developing next-generation data-driven wireless technologies that can adapt to the context. She served as TPC co-chair of the IEEE INFOCOM DeepWireless 2025-2026 Workshop and guest editor of npj Wireless Technology. She has received a Best Paper Award at IEEE INFOCOM 2025 and IEEE INFOCOM 2026. She is a 2023 Fulbright-Schuman alumna and is a recipient of an MSCA Global Postdoctoral Fellowship funded by the European Commission under the Horizon Europe scheme (2026–2029).