Álvaro Valcarce
Nokia Bell Labs, France
Title: Architecting the AI-RAN: Protocol Stack Evolution and Research Challenges for 6G

Abstract:
Radio Access Network (RAN) intelligence has historically been “compiled” into protocols: fixed rules, engineered interfaces, and long standardization cycles that trade adaptability for interoperability. As 6G targets extreme scale in services and devices, this waterfall process becomes slow and costly, and it struggles to capture site- and context-specific optimality. This lecture will frame protocol elements as learnable decision-making systems. Instead of handcrafting every policy and signaling convention, data-driven methods can train radio nodes to infer resource-allocation strategies and internalize protocol behaviors from interaction and traces (while remaining standards-compliant). The perspective aligns with the AI-RAN shift toward shared compute in the RAN and continuous improvement of control loops. The timing is reinforced by ecosystem moves: NVIDIA and Nokia announced an AI-RAN partnership, including a USD 1B NVIDIA investment in Nokia, to accelerate AI-native 6G RAN products on NVIDIA platforms. The AI-RAN Alliance is coordinating industry work around “AI-for-RAN”, “AI-and-RAN” and “AI-on-RAN” to benchmark, validate and operationalize AI-native RAN architectures at scale. This talk will highlight the research challenges of learning-based RAN protocols and will close with a brief look at how AI agents can automate parts of the wireless research workflow.

Short Bio:
Dr. Alvaro Valcarce is the Head of the AI for Wireless Research Department at Nokia Bell Labs, France, within the Radio Systems Research Lab. He leads a specialized team pioneering the AI-native Air Interface and the evolution of AI-RAN for 6G systems. His research focuses on the integration of advanced ML techniques (including Reinforcement Learning, Joint Embedding Predictive Architecture (JEPA), and Neuromorphic Computing) into radio resource management and protocol design. Dr. Valcarce holds a Master's degree in Telecommunications Engineering from the University of Vigo, Spain, and a Ph.D. in Computational Electromagnetics from the University of Bedfordshire, U.K., where his research focused on finite-difference time-domain (FDTD) modeling for large-scale radio channels. Recently, he served as the Chair of Industry-Academia Panels at the 2025 IEEE International Conference on Machine Learning for Communication and Networking (ICMLCN) and recently delivered a keynote speech at the Workshop on AI/ML for Next-Generation Wireless at NeurIPS 2025 in San Diego. Dr. Valcarce has co-authored numerous IEEE journal articles and book chapters on Radio Layers 2 and 3 and is a prolific inventor with a significant portfolio of patents in wireless optimization and AI-driven telecommunication systems.