AI-Native Receivers: Testing, Lifelong Adaptation, and Practical Deployment in Cellular Networks

Abstract

AI-native wireless receivers have shown impressive bit error rate (BER) gains over classical designs. Yet real-world adoption remains slow due to concerns about failure risk and operational uncertainty once models leave the lab. This talk presents an end-to-end view of how AI-native receivers can be made deployable around three practical pillars. First, we address testing before deployment using an AI-guided test generation framework that benchmarks pre-trained AI-native receivers without requiring exhaustive enumeration of channel and environment conditions. Second, we tackle continuous adaptation after deployment through a measurement–recovery framework that detects distribution shifts from pilots (e.g., changes in channel profile, speed, delay spread) and triggers finite retraining spurts only when needed. Our approach decides whether retraining is warranted even as deployment conditions drift. Third, we explore coexistence with classical receivers by switching between AI-native and classical decoding with a viewpoint of minimizing energy consumption when specific types of traffic are detected in the network.  Together, these three components outline a deployable blueprint for AI-native receivers that are testable before launch, self-correcting in the field, and interoperable with classical stacks.

Bio

Kaushik Chowdhury holds the Ernest Cockrell, Jr., Memorial Chair in the Electrical and Computer Engineering department at The University of Texas at Austin, USA. His research spans applied machine learning for wireless systems, including deep learning for spectrum sensing and sharing, RF cybersecurity, multimodal sensor fusion and digital twins. He is a Fellow of the IEEE. His work has been recognized with major honors, including being a finalist for the 2023 US Blavatnik National Awards for Young Scientists, the U.S. Presidential Early Career Award for Scientists and Engineers (PECASE), the DARPA Young Faculty Award, the ONR Director of Research Early Career Award, and the NSF CAREER Award.