Title: In-Radio Computing
Abstract: A radio is among the most sophisticated dynamical systems humans have engineered, yet we largely treat it as a conduit for signals. What if a radio could instead compute directly within the spectrum it observes, where broadband electromagnetic signals inherently reside? This question is particularly compelling given that modern wireless systems are defined by spectral complexity, while most computation remains grounded in time-domain abstractions. In this talk, we will see that this form of in-radio compute becomes possible by repurposing nonlinearity and poor quality-factors in CMOS traditionally viewed as limitations — into computational resources in a Microwave Neural Network (MNN) of multi-gigahertz frequency modes. This emerging research direction suggests a new model of fast, clockless computation driven directly by device physics. Preliminary experiments already demonstrate that MNN dynamics can be programmed to discriminate among wireless modulation formats and encode information through controllable statistical signatures and even emulate some digital functionalities. These programmable behaviors originate from interactions within tightly wound microwave waveguides, extending circuit functionality beyond conventional roles such as filters or delay lines for beam steering in wireless communication. This direction suggests the possibility of low-power circuits that may one day observe spectra across hundreds of gigahertz while actively extracting information and performing decision-making through their intrinsic dynamics in real time.
Bio: Bal Govind is a final-year PhD candidate in Electrical Engineering at Cornell University and a Knight Graduate Fellow at the Kavli Institute for Nanoscale Science at Cornell. His doctoral research focuses on the Microwave Neural Network (MNN), a high-speed physical computing circuit that combines millimeter-wave circuits and machine learning. His work explores broadband computation, RF signal processing, and hardware-efficient inference. He is also interested in the design of compact beam-steering components for wireless communication systems. Beyond microwave circuits, his research interests have included ultrasonic imaging and sensing.
Bal received master’s degrees in both Electrical Engineering (2021) and Mechanical Engineering (2018) from North Carolina State University. Prior to Cornell, he worked as a radio-frequency integrated circuit engineer at Texas Instruments and as a MEMS prototyping engineer at TDK InvenSense.
Host: Harish Krishnaswamy