Khalifa University, GSMA and AMD Launch AI Tool for Telecom Operators

Khalifa University of Science and Technology,  GSM Association (GSMA)  and AMD  are launching an operator evaluation initiative for RFGPT during MWC Doha 2026, opening access for telecom operators to explore its capabilities and assess its relevance to their radio frequency analysis needs.

The collaboration brings together Khalifa University’s research expertise in AI and wireless communications, GSMA’s global mobile industry ecosystem, and advanced computing capabilities from AMD to evaluate how foundation models can address emerging challenges in modern telecommunications networks.

Developed by Khalifa University’s Digital Future Institute, RFGPT is a Radio Frequency (RF) foundation model that combines advanced wireless signal understanding with natural-language interaction. The model allows users to ask questions about observed wireless signals, identify their characteristics, and interpret complex scenes containing multiple signals.

By translating highly technical radio frequency data into faster and more accessible insights, RFGPT has the potential to support operator functions such as radio access network (RAN) optimization, network planning, spectrum management, and interference analysis. Hosted on AMD compute and supported by GSMA, the model will be available to telecom operators to explore its capabilities and evaluate relevant use cases across network operations, planning and performance management.

RFGPT supports six evaluation areas. The model can identify wireless technologies including Wi-Fi, Bluetooth, LTE, 5G New Radio (NR), UMTS and DVB-S2, while distinguishing between uplink and downlink transmissions where applicable. It can also recognize modulation schemes and analyze complex wideband environments containing multiple simultaneous signals.

In addition, the model can assess signal coexistence by identifying time-frequency overlap between transmissions and estimating the strength of interference or signal interaction. It can estimate wireless activity levels by analysing WLAN and NR user densities within a given observation window, providing operators a better understanding of network utilization patterns. These capabilities could support more effective spectrum management, interference mitigation and capacity planning.

Reducing the time and specialist expertise typically required to interpret complex radio frequency data, RFGPT can also interpret key 5G NR network parameters, including subcarrier spacing, synchronization signal block (SSB) patterns, channel state information reference signal (CSI-RS) configurations and sounding reference signal (SRS) structures. This could enable engineers and network teams to access and act on radio frequency insights more quickly through natural-language interaction.

“Radio frequency foundation models offer a new way to interpret the wireless environment, combining signal understanding with natural-language reasoning. With RFGPT, we want to give operators an opportunity to explore this approach in practice and help shape its development around the challenges they encounter in their networks. The initiative also reflects Khalifa University’s commitment to advancing AI-driven communications technologies and accelerating the adoption of research innovations that address real-world challenges facing mobile network operators.”

– Professor Merouane Debbah, Senior Director, Digital Future Institute, Khalifa University

“Operators need opportunities to test emerging AI capabilities against concrete use cases. This collaboration provides a focused evaluation window for RFGPT, supported by API access on AMD compute. The feedback will help us understand where the model can deliver value and what is needed to take it further.”

– Louis Powell, Director of AI Technologies, GSMA

“As wireless networks grow more complex, operators need new ways to understand their radio environments and make informed decisions. Our collaboration with Khalifa University and GSMA gives operators access to RFGPT on AMD compute to evaluate how AI could simplify RF analysis and support more effective network planning, spectrum management and interference mitigation.”

– Philip Guido, Executive Vice President and Chief Commercial Officer, AMD

Interested operators are invited to submit their use cases. The research and hosting teams will coordinate the evaluation scope and access arrangements, with participating teams invited to share feedback on performance, usability, and relevance to their applications.